]>
Commit | Line | Data |
---|---|---|
db9ecf05 | 1 | /* SPDX-License-Identifier: LGPL-2.1-or-later */ |
15ae422b LP |
2 | |
3 | #include <errno.h> | |
e08f94ac | 4 | #include <linux/loop.h> |
07630cea | 5 | #include <sched.h> |
15ae422b | 6 | #include <stdio.h> |
07630cea | 7 | #include <sys/mount.h> |
07630cea | 8 | #include <unistd.h> |
25e870b5 | 9 | #include <linux/fs.h> |
15ae422b | 10 | |
b5efdb8a | 11 | #include "alloc-util.h" |
10404d52 | 12 | #include "base-filesystem.h" |
7f112f50 | 13 | #include "dev-setup.h" |
3ffd4af2 | 14 | #include "fd-util.h" |
e5f10caf | 15 | #include "format-util.h" |
d944dc95 | 16 | #include "fs-util.h" |
e908468b | 17 | #include "label.h" |
b3d13314 | 18 | #include "list.h" |
915e6d16 | 19 | #include "loop-util.h" |
07630cea | 20 | #include "loopback-setup.h" |
07630cea | 21 | #include "mkdir.h" |
4349cd7c | 22 | #include "mount-util.h" |
049af8ad | 23 | #include "mountpoint-util.h" |
0cb8e3d1 | 24 | #include "namespace-util.h" |
3ffd4af2 | 25 | #include "namespace.h" |
d8b4d14d | 26 | #include "nulstr-util.h" |
07630cea | 27 | #include "path-util.h" |
d7b8eec7 | 28 | #include "selinux-util.h" |
2583fbea | 29 | #include "socket-util.h" |
760877e9 | 30 | #include "sort-util.h" |
36ce7110 | 31 | #include "stat-util.h" |
8b43440b | 32 | #include "string-table.h" |
07630cea LP |
33 | #include "string-util.h" |
34 | #include "strv.h" | |
a652f050 | 35 | #include "tmpfile-util.h" |
affb60b1 | 36 | #include "umask-util.h" |
ee104e11 | 37 | #include "user-util.h" |
15ae422b | 38 | |
737ba3c8 | 39 | #define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC) |
40 | ||
c17ec25e | 41 | typedef enum MountMode { |
15ae422b LP |
42 | /* This is ordered by priority! */ |
43 | INACCESSIBLE, | |
b3d13314 | 44 | MOUNT_IMAGES, |
d2d6c096 LP |
45 | BIND_MOUNT, |
46 | BIND_MOUNT_RECURSIVE, | |
ac0930c8 | 47 | PRIVATE_TMP, |
56a13a49 | 48 | PRIVATE_TMP_READONLY, |
7f112f50 | 49 | PRIVATE_DEV, |
5d997827 | 50 | BIND_DEV, |
6c47cd7d | 51 | EMPTY_DIR, |
5d997827 LP |
52 | SYSFS, |
53 | PROCFS, | |
94293d65 | 54 | RUN, |
5d997827 | 55 | READONLY, |
59eeb84b | 56 | READWRITE, |
ddc155b2 TM |
57 | NOEXEC, |
58 | EXEC, | |
2abd4e38 | 59 | TMPFS, |
1e05071d | 60 | READWRITE_IMPLICIT, /* Should have the lowest priority. */ |
5beb8688 | 61 | _MOUNT_MODE_MAX, |
c17ec25e | 62 | } MountMode; |
15ae422b | 63 | |
34de407a | 64 | typedef struct MountEntry { |
5327c910 | 65 | const char *path_const; /* Memory allocated on stack or static */ |
cfbeb4ef | 66 | MountMode mode:5; |
5327c910 LP |
67 | bool ignore:1; /* Ignore if path does not exist? */ |
68 | bool has_prefix:1; /* Already is prefixed by the root dir? */ | |
cfbeb4ef | 69 | bool read_only:1; /* Shall this mount point be read-only? */ |
9ce4e4b0 | 70 | bool nosuid:1; /* Shall set MS_NOSUID on the mount itself */ |
ddc155b2 TM |
71 | bool noexec:1; /* Shall set MS_NOEXEC on the mount itself */ |
72 | bool exec:1; /* Shall clear MS_NOEXEC on the mount itself */ | |
088696fe | 73 | bool applied:1; /* Already applied */ |
55fe7432 | 74 | char *path_malloc; /* Use this instead of 'path_const' if we had to allocate memory */ |
b3d13314 | 75 | const char *source_const; /* The source path, for bind mounts or images */ |
d2d6c096 | 76 | char *source_malloc; |
2abd4e38 YW |
77 | const char *options_const;/* Mount options for tmpfs */ |
78 | char *options_malloc; | |
79 | unsigned long flags; /* Mount flags used by EMPTY_DIR and TMPFS. Do not include MS_RDONLY here, but please use read_only. */ | |
088696fe | 80 | unsigned n_followed; |
427353f6 | 81 | LIST_HEAD(MountOptions, image_options); |
34de407a | 82 | } MountEntry; |
15ae422b | 83 | |
94293d65 | 84 | /* If MountAPIVFS= is used, let's mount /sys, /proc, /dev and /run into the it, but only as a fallback if the user hasn't mounted |
3fe91079 | 85 | * something there already. These mounts are hence overridden by any other explicitly configured mounts. */ |
5d997827 LP |
86 | static const MountEntry apivfs_table[] = { |
87 | { "/proc", PROCFS, false }, | |
88 | { "/dev", BIND_DEV, false }, | |
89 | { "/sys", SYSFS, false }, | |
94293d65 | 90 | { "/run", RUN, false, .options_const = "mode=755" TMPFS_LIMITS_RUN, .flags = MS_NOSUID|MS_NODEV|MS_STRICTATIME }, |
5d997827 | 91 | }; |
f471b2af | 92 | |
11a30cec | 93 | /* ProtectKernelTunables= option and the related filesystem APIs */ |
34de407a | 94 | static const MountEntry protect_kernel_tunables_table[] = { |
1e05071d YW |
95 | { "/proc/acpi", READONLY, true }, |
96 | { "/proc/apm", READONLY, true }, /* Obsolete API, there's no point in permitting access to this, ever */ | |
97 | { "/proc/asound", READONLY, true }, | |
98 | { "/proc/bus", READONLY, true }, | |
99 | { "/proc/fs", READONLY, true }, | |
100 | { "/proc/irq", READONLY, true }, | |
101 | { "/proc/kallsyms", INACCESSIBLE, true }, | |
102 | { "/proc/kcore", INACCESSIBLE, true }, | |
103 | { "/proc/latency_stats", READONLY, true }, | |
104 | { "/proc/mtrr", READONLY, true }, | |
105 | { "/proc/scsi", READONLY, true }, | |
4e399953 | 106 | { "/proc/sys", READONLY, true }, |
1e05071d YW |
107 | { "/proc/sysrq-trigger", READONLY, true }, |
108 | { "/proc/timer_stats", READONLY, true }, | |
109 | { "/sys", READONLY, false }, | |
110 | { "/sys/fs/bpf", READONLY, true }, | |
111 | { "/sys/fs/cgroup", READWRITE_IMPLICIT, false }, /* READONLY is set by ProtectControlGroups= option */ | |
112 | { "/sys/fs/selinux", READWRITE_IMPLICIT, true }, | |
113 | { "/sys/kernel/debug", READONLY, true }, | |
114 | { "/sys/kernel/tracing", READONLY, true }, | |
11a30cec DH |
115 | }; |
116 | ||
c575770b | 117 | /* ProtectKernelModules= option */ |
34de407a | 118 | static const MountEntry protect_kernel_modules_table[] = { |
349cc4a5 | 119 | #if HAVE_SPLIT_USR |
c6232fb0 | 120 | { "/lib/modules", INACCESSIBLE, true }, |
c575770b | 121 | #endif |
c6232fb0 | 122 | { "/usr/lib/modules", INACCESSIBLE, true }, |
c575770b DH |
123 | }; |
124 | ||
94a7b275 KK |
125 | /* ProtectKernelLogs= option */ |
126 | static const MountEntry protect_kernel_logs_table[] = { | |
127 | { "/proc/kmsg", INACCESSIBLE, true }, | |
128 | { "/dev/kmsg", INACCESSIBLE, true }, | |
129 | }; | |
130 | ||
b6c432ca DH |
131 | /* |
132 | * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of | |
133 | * system should be protected by ProtectSystem= | |
134 | */ | |
34de407a | 135 | static const MountEntry protect_home_read_only_table[] = { |
c6232fb0 LP |
136 | { "/home", READONLY, true }, |
137 | { "/run/user", READONLY, true }, | |
138 | { "/root", READONLY, true }, | |
b6c432ca DH |
139 | }; |
140 | ||
e4da7d8c YW |
141 | /* ProtectHome=tmpfs table */ |
142 | static const MountEntry protect_home_tmpfs_table[] = { | |
7d85383e TM |
143 | { "/home", TMPFS, true, .read_only = true, .options_const = "mode=0755" TMPFS_LIMITS_EMPTY_OR_ALMOST, .flags = MS_NODEV|MS_STRICTATIME }, |
144 | { "/run/user", TMPFS, true, .read_only = true, .options_const = "mode=0755" TMPFS_LIMITS_EMPTY_OR_ALMOST, .flags = MS_NODEV|MS_STRICTATIME }, | |
145 | { "/root", TMPFS, true, .read_only = true, .options_const = "mode=0700" TMPFS_LIMITS_EMPTY_OR_ALMOST, .flags = MS_NODEV|MS_STRICTATIME }, | |
e4da7d8c YW |
146 | }; |
147 | ||
b6c432ca | 148 | /* ProtectHome=yes table */ |
34de407a | 149 | static const MountEntry protect_home_yes_table[] = { |
c6232fb0 LP |
150 | { "/home", INACCESSIBLE, true }, |
151 | { "/run/user", INACCESSIBLE, true }, | |
152 | { "/root", INACCESSIBLE, true }, | |
b6c432ca DH |
153 | }; |
154 | ||
f471b2af | 155 | /* ProtectSystem=yes table */ |
34de407a | 156 | static const MountEntry protect_system_yes_table[] = { |
c6232fb0 LP |
157 | { "/usr", READONLY, false }, |
158 | { "/boot", READONLY, true }, | |
159 | { "/efi", READONLY, true }, | |
7486f305 AB |
160 | #if HAVE_SPLIT_USR |
161 | { "/lib", READONLY, true }, | |
162 | { "/lib64", READONLY, true }, | |
163 | { "/bin", READONLY, true }, | |
671f0f8d | 164 | # if HAVE_SPLIT_BIN |
7486f305 | 165 | { "/sbin", READONLY, true }, |
671f0f8d | 166 | # endif |
7486f305 | 167 | #endif |
f471b2af DH |
168 | }; |
169 | ||
170 | /* ProtectSystem=full includes ProtectSystem=yes */ | |
34de407a | 171 | static const MountEntry protect_system_full_table[] = { |
c6232fb0 LP |
172 | { "/usr", READONLY, false }, |
173 | { "/boot", READONLY, true }, | |
174 | { "/efi", READONLY, true }, | |
175 | { "/etc", READONLY, false }, | |
7486f305 AB |
176 | #if HAVE_SPLIT_USR |
177 | { "/lib", READONLY, true }, | |
178 | { "/lib64", READONLY, true }, | |
179 | { "/bin", READONLY, true }, | |
671f0f8d | 180 | # if HAVE_SPLIT_BIN |
7486f305 | 181 | { "/sbin", READONLY, true }, |
671f0f8d | 182 | # endif |
7486f305 | 183 | #endif |
f471b2af DH |
184 | }; |
185 | ||
186 | /* | |
187 | * ProtectSystem=strict table. In this strict mode, we mount everything | |
188 | * read-only, except for /proc, /dev, /sys which are the kernel API VFS, | |
189 | * which are left writable, but PrivateDevices= + ProtectKernelTunables= | |
190 | * protect those, and these options should be fully orthogonal. | |
191 | * (And of course /home and friends are also left writable, as ProtectHome= | |
192 | * shall manage those, orthogonally). | |
193 | */ | |
34de407a | 194 | static const MountEntry protect_system_strict_table[] = { |
1e05071d YW |
195 | { "/", READONLY, false }, |
196 | { "/proc", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */ | |
197 | { "/sys", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */ | |
198 | { "/dev", READWRITE_IMPLICIT, false }, /* PrivateDevices= */ | |
199 | { "/home", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
200 | { "/run/user", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
201 | { "/root", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
f471b2af DH |
202 | }; |
203 | ||
5beb8688 YW |
204 | static const char * const mount_mode_table[_MOUNT_MODE_MAX] = { |
205 | [INACCESSIBLE] = "inaccessible", | |
206 | [BIND_MOUNT] = "bind", | |
207 | [BIND_MOUNT_RECURSIVE] = "rbind", | |
208 | [PRIVATE_TMP] = "private-tmp", | |
209 | [PRIVATE_DEV] = "private-dev", | |
210 | [BIND_DEV] = "bind-dev", | |
211 | [EMPTY_DIR] = "empty", | |
212 | [SYSFS] = "sysfs", | |
213 | [PROCFS] = "procfs", | |
214 | [READONLY] = "read-only", | |
215 | [READWRITE] = "read-write", | |
216 | [TMPFS] = "tmpfs", | |
b3d13314 | 217 | [MOUNT_IMAGES] = "mount-images", |
5beb8688 | 218 | [READWRITE_IMPLICIT] = "rw-implicit", |
ddc155b2 TM |
219 | [EXEC] = "exec", |
220 | [NOEXEC] = "noexec", | |
5beb8688 YW |
221 | }; |
222 | ||
223 | DEFINE_PRIVATE_STRING_TABLE_LOOKUP_TO_STRING(mount_mode, MountMode); | |
224 | ||
34de407a | 225 | static const char *mount_entry_path(const MountEntry *p) { |
f0a4feb0 DH |
226 | assert(p); |
227 | ||
5327c910 LP |
228 | /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that, |
229 | * otherwise the stack/static ->path field is returned. */ | |
f0a4feb0 | 230 | |
5327c910 | 231 | return p->path_malloc ?: p->path_const; |
f0a4feb0 DH |
232 | } |
233 | ||
34de407a | 234 | static bool mount_entry_read_only(const MountEntry *p) { |
cfbeb4ef LP |
235 | assert(p); |
236 | ||
56a13a49 | 237 | return p->read_only || IN_SET(p->mode, READONLY, INACCESSIBLE, PRIVATE_TMP_READONLY); |
cfbeb4ef LP |
238 | } |
239 | ||
ddc155b2 TM |
240 | static bool mount_entry_noexec(const MountEntry *p) { |
241 | assert(p); | |
242 | ||
243 | return p->noexec || IN_SET(p->mode, NOEXEC, INACCESSIBLE, SYSFS, PROCFS); | |
244 | } | |
245 | ||
246 | static bool mount_entry_exec(const MountEntry *p) { | |
247 | assert(p); | |
248 | ||
249 | return p->exec || p->mode == EXEC; | |
250 | } | |
251 | ||
d2d6c096 LP |
252 | static const char *mount_entry_source(const MountEntry *p) { |
253 | assert(p); | |
254 | ||
255 | return p->source_malloc ?: p->source_const; | |
256 | } | |
257 | ||
2abd4e38 YW |
258 | static const char *mount_entry_options(const MountEntry *p) { |
259 | assert(p); | |
260 | ||
261 | return p->options_malloc ?: p->options_const; | |
262 | } | |
263 | ||
1eb7e08e LP |
264 | static void mount_entry_done(MountEntry *p) { |
265 | assert(p); | |
266 | ||
267 | p->path_malloc = mfree(p->path_malloc); | |
268 | p->source_malloc = mfree(p->source_malloc); | |
2abd4e38 | 269 | p->options_malloc = mfree(p->options_malloc); |
427353f6 | 270 | p->image_options = mount_options_free_all(p->image_options); |
1eb7e08e LP |
271 | } |
272 | ||
d18aff04 | 273 | static int append_access_mounts(MountEntry **p, char **strv, MountMode mode, bool forcibly_require_prefix) { |
15ae422b LP |
274 | char **i; |
275 | ||
613b411c LP |
276 | assert(p); |
277 | ||
1e05071d | 278 | /* Adds a list of user-supplied READWRITE/READWRITE_IMPLICIT/READONLY/INACCESSIBLE entries */ |
5327c910 | 279 | |
15ae422b | 280 | STRV_FOREACH(i, strv) { |
5327c910 LP |
281 | bool ignore = false, needs_prefix = false; |
282 | const char *e = *i; | |
15ae422b | 283 | |
5327c910 LP |
284 | /* Look for any prefixes */ |
285 | if (startswith(e, "-")) { | |
286 | e++; | |
9c94d52e | 287 | ignore = true; |
ea92ae33 | 288 | } |
5327c910 LP |
289 | if (startswith(e, "+")) { |
290 | e++; | |
291 | needs_prefix = true; | |
292 | } | |
ea92ae33 | 293 | |
baaa35ad ZJS |
294 | if (!path_is_absolute(e)) |
295 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
296 | "Path is not absolute: %s", e); | |
15ae422b | 297 | |
34de407a | 298 | *((*p)++) = (MountEntry) { |
5327c910 LP |
299 | .path_const = e, |
300 | .mode = mode, | |
301 | .ignore = ignore, | |
d18aff04 | 302 | .has_prefix = !needs_prefix && !forcibly_require_prefix, |
5327c910 | 303 | }; |
15ae422b LP |
304 | } |
305 | ||
306 | return 0; | |
307 | } | |
308 | ||
6c47cd7d LP |
309 | static int append_empty_dir_mounts(MountEntry **p, char **strv) { |
310 | char **i; | |
311 | ||
312 | assert(p); | |
313 | ||
314 | /* Adds tmpfs mounts to provide readable but empty directories. This is primarily used to implement the | |
315 | * "/private/" boundary directories for DynamicUser=1. */ | |
316 | ||
317 | STRV_FOREACH(i, strv) { | |
318 | ||
319 | *((*p)++) = (MountEntry) { | |
320 | .path_const = *i, | |
321 | .mode = EMPTY_DIR, | |
322 | .ignore = false, | |
6c47cd7d | 323 | .read_only = true, |
7d85383e | 324 | .options_const = "mode=755" TMPFS_LIMITS_EMPTY_OR_ALMOST, |
2abd4e38 | 325 | .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, |
6c47cd7d LP |
326 | }; |
327 | } | |
328 | ||
329 | return 0; | |
330 | } | |
331 | ||
da6053d0 | 332 | static int append_bind_mounts(MountEntry **p, const BindMount *binds, size_t n) { |
d2d6c096 LP |
333 | assert(p); |
334 | ||
fe96c0f8 | 335 | for (size_t i = 0; i < n; i++) { |
d2d6c096 LP |
336 | const BindMount *b = binds + i; |
337 | ||
338 | *((*p)++) = (MountEntry) { | |
339 | .path_const = b->destination, | |
340 | .mode = b->recursive ? BIND_MOUNT_RECURSIVE : BIND_MOUNT, | |
341 | .read_only = b->read_only, | |
9ce4e4b0 | 342 | .nosuid = b->nosuid, |
d2d6c096 | 343 | .source_const = b->source, |
4ca763a9 | 344 | .ignore = b->ignore_enoent, |
d2d6c096 LP |
345 | }; |
346 | } | |
347 | ||
348 | return 0; | |
349 | } | |
350 | ||
b3d13314 LB |
351 | static int append_mount_images(MountEntry **p, const MountImage *mount_images, size_t n) { |
352 | assert(p); | |
353 | ||
354 | for (size_t i = 0; i < n; i++) { | |
355 | const MountImage *m = mount_images + i; | |
356 | ||
357 | *((*p)++) = (MountEntry) { | |
358 | .path_const = m->destination, | |
359 | .mode = MOUNT_IMAGES, | |
360 | .source_const = m->source, | |
427353f6 | 361 | .image_options = m->mount_options, |
b3d13314 LB |
362 | .ignore = m->ignore_enoent, |
363 | }; | |
364 | } | |
365 | ||
366 | return 0; | |
367 | } | |
368 | ||
da6053d0 | 369 | static int append_tmpfs_mounts(MountEntry **p, const TemporaryFileSystem *tmpfs, size_t n) { |
2abd4e38 YW |
370 | assert(p); |
371 | ||
b67ec8e5 | 372 | for (size_t i = 0; i < n; i++) { |
2abd4e38 YW |
373 | const TemporaryFileSystem *t = tmpfs + i; |
374 | _cleanup_free_ char *o = NULL, *str = NULL; | |
ad8e66dc | 375 | unsigned long flags; |
2abd4e38 | 376 | bool ro = false; |
b67ec8e5 | 377 | int r; |
2abd4e38 | 378 | |
baaa35ad ZJS |
379 | if (!path_is_absolute(t->path)) |
380 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
381 | "Path is not absolute: %s", | |
382 | t->path); | |
2abd4e38 | 383 | |
b67ec8e5 | 384 | str = strjoin("mode=0755" NESTED_TMPFS_LIMITS ",", t->options); |
ad8e66dc AJ |
385 | if (!str) |
386 | return -ENOMEM; | |
2abd4e38 | 387 | |
ad8e66dc AJ |
388 | r = mount_option_mangle(str, MS_NODEV|MS_STRICTATIME, &flags, &o); |
389 | if (r < 0) | |
390 | return log_debug_errno(r, "Failed to parse mount option '%s': %m", str); | |
2abd4e38 | 391 | |
ad8e66dc AJ |
392 | ro = flags & MS_RDONLY; |
393 | if (ro) | |
394 | flags ^= MS_RDONLY; | |
2abd4e38 YW |
395 | |
396 | *((*p)++) = (MountEntry) { | |
397 | .path_const = t->path, | |
398 | .mode = TMPFS, | |
399 | .read_only = ro, | |
ad8e66dc | 400 | .options_malloc = TAKE_PTR(o), |
2abd4e38 YW |
401 | .flags = flags, |
402 | }; | |
2abd4e38 YW |
403 | } |
404 | ||
405 | return 0; | |
406 | } | |
407 | ||
da6053d0 | 408 | static int append_static_mounts(MountEntry **p, const MountEntry *mounts, size_t n, bool ignore_protect) { |
11a30cec | 409 | assert(p); |
f471b2af | 410 | assert(mounts); |
11a30cec | 411 | |
5327c910 | 412 | /* Adds a list of static pre-defined entries */ |
f471b2af | 413 | |
fe96c0f8 | 414 | for (size_t i = 0; i < n; i++) |
34de407a LP |
415 | *((*p)++) = (MountEntry) { |
416 | .path_const = mount_entry_path(mounts+i), | |
5327c910 LP |
417 | .mode = mounts[i].mode, |
418 | .ignore = mounts[i].ignore || ignore_protect, | |
419 | }; | |
f471b2af DH |
420 | |
421 | return 0; | |
422 | } | |
423 | ||
34de407a | 424 | static int append_protect_home(MountEntry **p, ProtectHome protect_home, bool ignore_protect) { |
c575770b DH |
425 | assert(p); |
426 | ||
5327c910 | 427 | switch (protect_home) { |
b6c432ca | 428 | |
5327c910 | 429 | case PROTECT_HOME_NO: |
b6c432ca DH |
430 | return 0; |
431 | ||
b6c432ca | 432 | case PROTECT_HOME_READ_ONLY: |
5327c910 LP |
433 | return append_static_mounts(p, protect_home_read_only_table, ELEMENTSOF(protect_home_read_only_table), ignore_protect); |
434 | ||
e4da7d8c YW |
435 | case PROTECT_HOME_TMPFS: |
436 | return append_static_mounts(p, protect_home_tmpfs_table, ELEMENTSOF(protect_home_tmpfs_table), ignore_protect); | |
437 | ||
b6c432ca | 438 | case PROTECT_HOME_YES: |
5327c910 LP |
439 | return append_static_mounts(p, protect_home_yes_table, ELEMENTSOF(protect_home_yes_table), ignore_protect); |
440 | ||
b6c432ca | 441 | default: |
5327c910 | 442 | assert_not_reached("Unexpected ProtectHome= value"); |
b6c432ca | 443 | } |
b6c432ca DH |
444 | } |
445 | ||
34de407a | 446 | static int append_protect_system(MountEntry **p, ProtectSystem protect_system, bool ignore_protect) { |
f471b2af DH |
447 | assert(p); |
448 | ||
5327c910 LP |
449 | switch (protect_system) { |
450 | ||
451 | case PROTECT_SYSTEM_NO: | |
f471b2af DH |
452 | return 0; |
453 | ||
f471b2af | 454 | case PROTECT_SYSTEM_STRICT: |
5327c910 LP |
455 | return append_static_mounts(p, protect_system_strict_table, ELEMENTSOF(protect_system_strict_table), ignore_protect); |
456 | ||
f471b2af | 457 | case PROTECT_SYSTEM_YES: |
5327c910 LP |
458 | return append_static_mounts(p, protect_system_yes_table, ELEMENTSOF(protect_system_yes_table), ignore_protect); |
459 | ||
f471b2af | 460 | case PROTECT_SYSTEM_FULL: |
5327c910 LP |
461 | return append_static_mounts(p, protect_system_full_table, ELEMENTSOF(protect_system_full_table), ignore_protect); |
462 | ||
f471b2af | 463 | default: |
5327c910 | 464 | assert_not_reached("Unexpected ProtectSystem= value"); |
f471b2af | 465 | } |
11a30cec DH |
466 | } |
467 | ||
93bab288 | 468 | static int mount_path_compare(const MountEntry *a, const MountEntry *b) { |
a0827e2b | 469 | int d; |
15ae422b | 470 | |
6ee1a919 | 471 | /* If the paths are not equal, then order prefixes first */ |
93bab288 | 472 | d = path_compare(mount_entry_path(a), mount_entry_path(b)); |
6ee1a919 LP |
473 | if (d != 0) |
474 | return d; | |
15ae422b | 475 | |
6ee1a919 | 476 | /* If the paths are equal, check the mode */ |
93bab288 | 477 | return CMP((int) a->mode, (int) b->mode); |
15ae422b LP |
478 | } |
479 | ||
da6053d0 | 480 | static int prefix_where_needed(MountEntry *m, size_t n, const char *root_directory) { |
4a756839 | 481 | /* Prefixes all paths in the bind mount table with the root directory if the entry needs that. */ |
5327c910 | 482 | |
fe96c0f8 SS |
483 | assert(m || n == 0); |
484 | ||
485 | for (size_t i = 0; i < n; i++) { | |
5327c910 LP |
486 | char *s; |
487 | ||
488 | if (m[i].has_prefix) | |
489 | continue; | |
490 | ||
c6134d3e | 491 | s = path_join(root_directory, mount_entry_path(m+i)); |
5327c910 LP |
492 | if (!s) |
493 | return -ENOMEM; | |
494 | ||
e282f51f | 495 | free_and_replace(m[i].path_malloc, s); |
5327c910 LP |
496 | m[i].has_prefix = true; |
497 | } | |
498 | ||
499 | return 0; | |
500 | } | |
501 | ||
da6053d0 | 502 | static void drop_duplicates(MountEntry *m, size_t *n) { |
34de407a | 503 | MountEntry *f, *t, *previous; |
15ae422b | 504 | |
c17ec25e | 505 | assert(m); |
15ae422b | 506 | assert(n); |
15ae422b | 507 | |
fe3c2583 LP |
508 | /* Drops duplicate entries. Expects that the array is properly ordered already. */ |
509 | ||
1d54cd5d | 510 | for (f = m, t = m, previous = NULL; f < m + *n; f++) { |
15ae422b | 511 | |
fe3c2583 | 512 | /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare() |
088696fe LP |
513 | * above. Note that we only drop duplicates that haven't been mounted yet. */ |
514 | if (previous && | |
515 | path_equal(mount_entry_path(f), mount_entry_path(previous)) && | |
516 | !f->applied && !previous->applied) { | |
5beb8688 | 517 | log_debug("%s (%s) is duplicate.", mount_entry_path(f), mount_mode_to_string(f->mode)); |
ddc155b2 TM |
518 | /* Propagate the flags to the remaining entry */ |
519 | previous->read_only = previous->read_only || mount_entry_read_only(f); | |
520 | previous->noexec = previous->noexec || mount_entry_noexec(f); | |
521 | previous->exec = previous->exec || mount_entry_exec(f); | |
1eb7e08e | 522 | mount_entry_done(f); |
15ae422b | 523 | continue; |
fe3c2583 | 524 | } |
15ae422b | 525 | |
e2d7c1a0 | 526 | *t = *f; |
15ae422b | 527 | previous = t; |
fe3c2583 LP |
528 | t++; |
529 | } | |
530 | ||
531 | *n = t - m; | |
532 | } | |
533 | ||
da6053d0 | 534 | static void drop_inaccessible(MountEntry *m, size_t *n) { |
34de407a | 535 | MountEntry *f, *t; |
fe3c2583 LP |
536 | const char *clear = NULL; |
537 | ||
538 | assert(m); | |
539 | assert(n); | |
540 | ||
541 | /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly | |
542 | * ordered already. */ | |
543 | ||
1d54cd5d | 544 | for (f = m, t = m; f < m + *n; f++) { |
fe3c2583 LP |
545 | |
546 | /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop | |
547 | * it, as inaccessible paths really should drop the entire subtree. */ | |
34de407a LP |
548 | if (clear && path_startswith(mount_entry_path(f), clear)) { |
549 | log_debug("%s is masked by %s.", mount_entry_path(f), clear); | |
1eb7e08e | 550 | mount_entry_done(f); |
fe3c2583 LP |
551 | continue; |
552 | } | |
15ae422b | 553 | |
34de407a | 554 | clear = f->mode == INACCESSIBLE ? mount_entry_path(f) : NULL; |
fe3c2583 LP |
555 | |
556 | *t = *f; | |
15ae422b LP |
557 | t++; |
558 | } | |
559 | ||
c17ec25e | 560 | *n = t - m; |
15ae422b LP |
561 | } |
562 | ||
da6053d0 | 563 | static void drop_nop(MountEntry *m, size_t *n) { |
34de407a | 564 | MountEntry *f, *t; |
7648a565 LP |
565 | |
566 | assert(m); | |
567 | assert(n); | |
568 | ||
569 | /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the | |
570 | * list is ordered by prefixes. */ | |
571 | ||
1d54cd5d | 572 | for (f = m, t = m; f < m + *n; f++) { |
7648a565 | 573 | |
1e05071d YW |
574 | /* Only suppress such subtrees for READONLY, READWRITE and READWRITE_IMPLICIT entries */ |
575 | if (IN_SET(f->mode, READONLY, READWRITE, READWRITE_IMPLICIT)) { | |
34de407a | 576 | MountEntry *p; |
7648a565 LP |
577 | bool found = false; |
578 | ||
579 | /* Now let's find the first parent of the entry we are looking at. */ | |
580 | for (p = t-1; p >= m; p--) { | |
34de407a | 581 | if (path_startswith(mount_entry_path(f), mount_entry_path(p))) { |
7648a565 LP |
582 | found = true; |
583 | break; | |
584 | } | |
585 | } | |
586 | ||
587 | /* We found it, let's see if it's the same mode, if so, we can drop this entry */ | |
588 | if (found && p->mode == f->mode) { | |
5beb8688 YW |
589 | log_debug("%s (%s) is made redundant by %s (%s)", |
590 | mount_entry_path(f), mount_mode_to_string(f->mode), | |
591 | mount_entry_path(p), mount_mode_to_string(p->mode)); | |
1eb7e08e | 592 | mount_entry_done(f); |
7648a565 LP |
593 | continue; |
594 | } | |
595 | } | |
596 | ||
597 | *t = *f; | |
598 | t++; | |
599 | } | |
600 | ||
601 | *n = t - m; | |
602 | } | |
603 | ||
da6053d0 | 604 | static void drop_outside_root(const char *root_directory, MountEntry *m, size_t *n) { |
34de407a | 605 | MountEntry *f, *t; |
cd2902c9 LP |
606 | |
607 | assert(m); | |
608 | assert(n); | |
609 | ||
1d54cd5d | 610 | /* Nothing to do */ |
cd2902c9 LP |
611 | if (!root_directory) |
612 | return; | |
613 | ||
614 | /* Drops all mounts that are outside of the root directory. */ | |
615 | ||
1d54cd5d | 616 | for (f = m, t = m; f < m + *n; f++) { |
cd2902c9 | 617 | |
34de407a LP |
618 | if (!path_startswith(mount_entry_path(f), root_directory)) { |
619 | log_debug("%s is outside of root directory.", mount_entry_path(f)); | |
1eb7e08e | 620 | mount_entry_done(f); |
cd2902c9 LP |
621 | continue; |
622 | } | |
623 | ||
624 | *t = *f; | |
625 | t++; | |
626 | } | |
627 | ||
628 | *n = t - m; | |
629 | } | |
630 | ||
b2a60844 LP |
631 | static int clone_device_node( |
632 | const char *d, | |
633 | const char *temporary_mount, | |
634 | bool *make_devnode) { | |
635 | ||
636 | _cleanup_free_ char *sl = NULL; | |
637 | const char *dn, *bn, *t; | |
b5e99f23 ДГ |
638 | struct stat st; |
639 | int r; | |
640 | ||
414b304b | 641 | if (stat(d, &st) < 0) { |
b2a60844 LP |
642 | if (errno == ENOENT) { |
643 | log_debug_errno(errno, "Device node '%s' to clone does not exist, ignoring.", d); | |
af984e13 | 644 | return -ENXIO; |
b2a60844 LP |
645 | } |
646 | ||
647 | return log_debug_errno(errno, "Failed to stat() device node '%s' to clone, ignoring: %m", d); | |
b5e99f23 ДГ |
648 | } |
649 | ||
650 | if (!S_ISBLK(st.st_mode) && | |
baaa35ad ZJS |
651 | !S_ISCHR(st.st_mode)) |
652 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
653 | "Device node '%s' to clone is not a device node, ignoring.", | |
654 | d); | |
b5e99f23 | 655 | |
6f7f3a33 | 656 | dn = strjoina(temporary_mount, d); |
b5e99f23 | 657 | |
b2a60844 | 658 | /* First, try to create device node properly */ |
16498617 CB |
659 | if (*make_devnode) { |
660 | mac_selinux_create_file_prepare(d, st.st_mode); | |
661 | r = mknod(dn, st.st_mode, st.st_rdev); | |
662 | mac_selinux_create_file_clear(); | |
b2a60844 LP |
663 | if (r >= 0) |
664 | goto add_symlink; | |
16498617 CB |
665 | if (errno != EPERM) |
666 | return log_debug_errno(errno, "mknod failed for %s: %m", d); | |
667 | ||
b2a60844 | 668 | /* This didn't work, let's not try this again for the next iterations. */ |
16498617 CB |
669 | *make_devnode = false; |
670 | } | |
671 | ||
2aed63f4 | 672 | /* We're about to fall back to bind-mounting the device |
1acf344d CG |
673 | * node. So create a dummy bind-mount target. |
674 | * Do not prepare device-node SELinux label (see issue 13762) */ | |
16498617 | 675 | r = mknod(dn, S_IFREG, 0); |
16498617 | 676 | if (r < 0 && errno != EEXIST) |
b2a60844 | 677 | return log_debug_errno(errno, "mknod() fallback failed for '%s': %m", d); |
16498617 | 678 | |
21935150 LP |
679 | /* Fallback to bind-mounting: The assumption here is that all used device nodes carry standard |
680 | * properties. Specifically, the devices nodes we bind-mount should either be owned by root:root or | |
681 | * root:tty (e.g. /dev/tty, /dev/ptmx) and should not carry ACLs. */ | |
682 | r = mount_nofollow_verbose(LOG_DEBUG, d, dn, NULL, MS_BIND, NULL); | |
683 | if (r < 0) | |
684 | return r; | |
b2a60844 LP |
685 | |
686 | add_symlink: | |
687 | bn = path_startswith(d, "/dev/"); | |
688 | if (!bn) | |
689 | return 0; | |
690 | ||
691 | /* Create symlinks like /dev/char/1:9 → ../urandom */ | |
cbc056c8 ZJS |
692 | if (asprintf(&sl, "%s/dev/%s/%u:%u", |
693 | temporary_mount, | |
694 | S_ISCHR(st.st_mode) ? "char" : "block", | |
695 | major(st.st_rdev), minor(st.st_rdev)) < 0) | |
b2a60844 LP |
696 | return log_oom(); |
697 | ||
698 | (void) mkdir_parents(sl, 0755); | |
699 | ||
700 | t = strjoina("../", bn); | |
b2a60844 | 701 | if (symlink(t, sl) < 0) |
2e4a4fae | 702 | log_debug_errno(errno, "Failed to symlink '%s' to '%s', ignoring: %m", t, sl); |
b5e99f23 | 703 | |
af984e13 | 704 | return 0; |
b5e99f23 ДГ |
705 | } |
706 | ||
5d997827 | 707 | static int mount_private_dev(MountEntry *m) { |
7f112f50 LP |
708 | static const char devnodes[] = |
709 | "/dev/null\0" | |
710 | "/dev/zero\0" | |
711 | "/dev/full\0" | |
712 | "/dev/random\0" | |
713 | "/dev/urandom\0" | |
714 | "/dev/tty\0"; | |
715 | ||
2b85f4e1 | 716 | char temporary_mount[] = "/tmp/namespace-dev-XXXXXX"; |
63cc4c31 | 717 | const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL; |
16498617 | 718 | bool can_mknod = true; |
7f112f50 LP |
719 | _cleanup_umask_ mode_t u; |
720 | int r; | |
721 | ||
722 | assert(m); | |
723 | ||
724 | u = umask(0000); | |
725 | ||
2b85f4e1 | 726 | if (!mkdtemp(temporary_mount)) |
2e4a4fae | 727 | return log_debug_errno(errno, "Failed to create temporary directory '%s': %m", temporary_mount); |
2b85f4e1 | 728 | |
63c372cb | 729 | dev = strjoina(temporary_mount, "/dev"); |
dc751688 | 730 | (void) mkdir(dev, 0755); |
21935150 LP |
731 | r = mount_nofollow_verbose(LOG_DEBUG, "tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755" TMPFS_LIMITS_DEV); |
732 | if (r < 0) | |
2b85f4e1 | 733 | goto fail; |
21935150 | 734 | |
c3151977 TM |
735 | r = label_fix_container(dev, "/dev", 0); |
736 | if (r < 0) { | |
737 | log_debug_errno(errno, "Failed to fix label of '%s' as /dev: %m", dev); | |
738 | goto fail; | |
739 | } | |
2b85f4e1 | 740 | |
63c372cb | 741 | devpts = strjoina(temporary_mount, "/dev/pts"); |
dc751688 | 742 | (void) mkdir(devpts, 0755); |
21935150 LP |
743 | r = mount_nofollow_verbose(LOG_DEBUG, "/dev/pts", devpts, NULL, MS_BIND, NULL); |
744 | if (r < 0) | |
2b85f4e1 | 745 | goto fail; |
2b85f4e1 | 746 | |
2e4a4fae YW |
747 | /* /dev/ptmx can either be a device node or a symlink to /dev/pts/ptmx. |
748 | * When /dev/ptmx a device node, /dev/pts/ptmx has 000 permissions making it inaccessible. | |
749 | * Thus, in that case make a clone. | |
750 | * In nspawn and other containers it will be a symlink, in that case make it a symlink. */ | |
36ce7110 | 751 | r = is_symlink("/dev/ptmx"); |
2e4a4fae YW |
752 | if (r < 0) { |
753 | log_debug_errno(r, "Failed to detect whether /dev/ptmx is a symlink or not: %m"); | |
3164e3cb | 754 | goto fail; |
2e4a4fae | 755 | } else if (r > 0) { |
414b304b ДГ |
756 | devptmx = strjoina(temporary_mount, "/dev/ptmx"); |
757 | if (symlink("pts/ptmx", devptmx) < 0) { | |
2e4a4fae | 758 | r = log_debug_errno(errno, "Failed to create a symlink '%s' to pts/ptmx: %m", devptmx); |
414b304b ДГ |
759 | goto fail; |
760 | } | |
761 | } else { | |
16498617 | 762 | r = clone_device_node("/dev/ptmx", temporary_mount, &can_mknod); |
152c475f LP |
763 | if (r < 0) |
764 | goto fail; | |
414b304b | 765 | } |
e06b6479 | 766 | |
63c372cb | 767 | devshm = strjoina(temporary_mount, "/dev/shm"); |
8d953682 | 768 | (void) mkdir(devshm, 0755); |
21935150 LP |
769 | r = mount_nofollow_verbose(LOG_DEBUG, "/dev/shm", devshm, NULL, MS_BIND, NULL); |
770 | if (r < 0) | |
2b85f4e1 | 771 | goto fail; |
2b85f4e1 | 772 | |
63c372cb | 773 | devmqueue = strjoina(temporary_mount, "/dev/mqueue"); |
dc751688 | 774 | (void) mkdir(devmqueue, 0755); |
21935150 | 775 | (void) mount_nofollow_verbose(LOG_DEBUG, "/dev/mqueue", devmqueue, NULL, MS_BIND, NULL); |
2b85f4e1 | 776 | |
63c372cb | 777 | devhugepages = strjoina(temporary_mount, "/dev/hugepages"); |
dc751688 | 778 | (void) mkdir(devhugepages, 0755); |
21935150 | 779 | (void) mount_nofollow_verbose(LOG_DEBUG, "/dev/hugepages", devhugepages, NULL, MS_BIND, NULL); |
2b85f4e1 | 780 | |
63c372cb | 781 | devlog = strjoina(temporary_mount, "/dev/log"); |
2e4a4fae YW |
782 | if (symlink("/run/systemd/journal/dev-log", devlog) < 0) |
783 | log_debug_errno(errno, "Failed to create a symlink '%s' to /run/systemd/journal/dev-log, ignoring: %m", devlog); | |
82d25240 | 784 | |
7f112f50 | 785 | NULSTR_FOREACH(d, devnodes) { |
16498617 | 786 | r = clone_device_node(d, temporary_mount, &can_mknod); |
37b22b3b | 787 | /* ENXIO means the *source* is not a device file, skip creation in that case */ |
af984e13 | 788 | if (r < 0 && r != -ENXIO) |
2b85f4e1 | 789 | goto fail; |
7f112f50 LP |
790 | } |
791 | ||
2e4a4fae YW |
792 | r = dev_setup(temporary_mount, UID_INVALID, GID_INVALID); |
793 | if (r < 0) | |
105a1a36 | 794 | log_debug_errno(r, "Failed to set up basic device tree at '%s', ignoring: %m", temporary_mount); |
7f112f50 | 795 | |
ee818b89 AC |
796 | /* Create the /dev directory if missing. It is more likely to be |
797 | * missing when the service is started with RootDirectory. This is | |
798 | * consistent with mount units creating the mount points when missing. | |
799 | */ | |
34de407a | 800 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
ee818b89 | 801 | |
9e5f8252 | 802 | /* Unmount everything in old /dev */ |
2e4a4fae YW |
803 | r = umount_recursive(mount_entry_path(m), 0); |
804 | if (r < 0) | |
805 | log_debug_errno(r, "Failed to unmount directories below '%s', ignoring: %m", mount_entry_path(m)); | |
806 | ||
21935150 LP |
807 | r = mount_nofollow_verbose(LOG_DEBUG, dev, mount_entry_path(m), NULL, MS_MOVE, NULL); |
808 | if (r < 0) | |
2b85f4e1 | 809 | goto fail; |
7f112f50 | 810 | |
1019a48f LP |
811 | (void) rmdir(dev); |
812 | (void) rmdir(temporary_mount); | |
7f112f50 | 813 | |
2b85f4e1 | 814 | return 0; |
7f112f50 | 815 | |
2b85f4e1 LP |
816 | fail: |
817 | if (devpts) | |
21935150 | 818 | (void) umount_verbose(LOG_DEBUG, devpts, UMOUNT_NOFOLLOW); |
7f112f50 | 819 | |
2b85f4e1 | 820 | if (devshm) |
21935150 | 821 | (void) umount_verbose(LOG_DEBUG, devshm, UMOUNT_NOFOLLOW); |
7f112f50 | 822 | |
2b85f4e1 | 823 | if (devhugepages) |
21935150 | 824 | (void) umount_verbose(LOG_DEBUG, devhugepages, UMOUNT_NOFOLLOW); |
7f112f50 | 825 | |
2b85f4e1 | 826 | if (devmqueue) |
21935150 | 827 | (void) umount_verbose(LOG_DEBUG, devmqueue, UMOUNT_NOFOLLOW); |
7f112f50 | 828 | |
21935150 | 829 | (void) umount_verbose(LOG_DEBUG, dev, UMOUNT_NOFOLLOW); |
1019a48f LP |
830 | (void) rmdir(dev); |
831 | (void) rmdir(temporary_mount); | |
7f112f50 | 832 | |
2b85f4e1 | 833 | return r; |
7f112f50 LP |
834 | } |
835 | ||
2a2969fd | 836 | static int mount_bind_dev(const MountEntry *m) { |
5d997827 LP |
837 | int r; |
838 | ||
839 | assert(m); | |
840 | ||
841 | /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's | |
842 | * /dev. This is only used when RootDirectory= is set. */ | |
843 | ||
645767d6 LP |
844 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
845 | ||
5d997827 LP |
846 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
847 | if (r < 0) | |
848 | return log_debug_errno(r, "Unable to determine whether /dev is already mounted: %m"); | |
849 | if (r > 0) /* make this a NOP if /dev is already a mount point */ | |
850 | return 0; | |
851 | ||
21935150 LP |
852 | r = mount_nofollow_verbose(LOG_DEBUG, "/dev", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL); |
853 | if (r < 0) | |
854 | return r; | |
5d997827 LP |
855 | |
856 | return 1; | |
857 | } | |
858 | ||
2a2969fd | 859 | static int mount_sysfs(const MountEntry *m) { |
5d997827 LP |
860 | int r; |
861 | ||
862 | assert(m); | |
863 | ||
645767d6 LP |
864 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
865 | ||
5d997827 LP |
866 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
867 | if (r < 0) | |
868 | return log_debug_errno(r, "Unable to determine whether /sys is already mounted: %m"); | |
869 | if (r > 0) /* make this a NOP if /sys is already a mount point */ | |
870 | return 0; | |
871 | ||
872 | /* Bind mount the host's version so that we get all child mounts of it, too. */ | |
21935150 LP |
873 | r = mount_nofollow_verbose(LOG_DEBUG, "/sys", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL); |
874 | if (r < 0) | |
875 | return r; | |
5d997827 LP |
876 | |
877 | return 1; | |
878 | } | |
879 | ||
4e399953 | 880 | static int mount_procfs(const MountEntry *m, const NamespaceInfo *ns_info) { |
61f8a7bd | 881 | _cleanup_free_ char *opts = NULL; |
4e399953 | 882 | const char *entry_path; |
61f8a7bd | 883 | int r, n; |
5d997827 LP |
884 | |
885 | assert(m); | |
4e399953 | 886 | assert(ns_info); |
5d997827 | 887 | |
4e399953 LP |
888 | if (ns_info->protect_proc != PROTECT_PROC_DEFAULT || |
889 | ns_info->proc_subset != PROC_SUBSET_ALL) { | |
4e399953 LP |
890 | |
891 | /* Starting with kernel 5.8 procfs' hidepid= logic is truly per-instance (previously it | |
892 | * pretended to be per-instance but actually was per-namespace), hence let's make use of it | |
893 | * if requested. To make sure this logic succeeds only on kernels where hidepid= is | |
894 | * per-instance, we'll exclusively use the textual value for hidepid=, since support was | |
895 | * added in the same commit: if it's supported it is thus also per-instance. */ | |
896 | ||
897 | opts = strjoin("hidepid=", | |
898 | ns_info->protect_proc == PROTECT_PROC_DEFAULT ? "off" : | |
899 | protect_proc_to_string(ns_info->protect_proc), | |
900 | ns_info->proc_subset == PROC_SUBSET_PID ? ",subset=pid" : ""); | |
901 | if (!opts) | |
902 | return -ENOMEM; | |
4e399953 LP |
903 | } |
904 | ||
61f8a7bd YW |
905 | entry_path = mount_entry_path(m); |
906 | (void) mkdir_p_label(entry_path, 0755); | |
907 | ||
908 | /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in | |
909 | * one. i.e we don't reuse existing mounts here under any condition, we want a new instance owned by | |
910 | * our user namespace and with our hidepid= settings applied. Hence, let's get rid of everything | |
911 | * mounted on /proc/ first. */ | |
912 | ||
913 | n = umount_recursive(entry_path, 0); | |
914 | ||
915 | r = mount_nofollow_verbose(LOG_DEBUG, "proc", entry_path, "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, opts); | |
916 | if (r == -EINVAL && opts) | |
917 | /* If this failed with EINVAL then this likely means the textual hidepid= stuff is | |
918 | * not supported by the kernel, and thus the per-instance hidepid= neither, which | |
919 | * means we really don't want to use it, since it would affect our host's /proc | |
920 | * mount. Hence let's gracefully fallback to a classic, unrestricted version. */ | |
921 | r = mount_nofollow_verbose(LOG_DEBUG, "proc", entry_path, "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL); | |
922 | if (r == -EPERM) { | |
24ee0f9d | 923 | /* When we do not have enough privileges to mount /proc, fallback to use existing /proc. */ |
61f8a7bd YW |
924 | |
925 | if (n > 0) | |
926 | /* /proc or some of sub-mounts are umounted in the above. Refuse incomplete tree. | |
927 | * Propagate the original error code returned by mount() in the above. */ | |
928 | return -EPERM; | |
929 | ||
930 | r = path_is_mount_point(entry_path, NULL, 0); | |
931 | if (r < 0) | |
932 | return log_debug_errno(r, "Unable to determine whether /proc is already mounted: %m"); | |
933 | if (r == 0) | |
934 | /* /proc is not mounted. Propagate the original error code. */ | |
935 | return -EPERM; | |
ad74f28a YW |
936 | } else if (r < 0) |
937 | return r; | |
5d997827 LP |
938 | |
939 | return 1; | |
940 | } | |
941 | ||
2abd4e38 | 942 | static int mount_tmpfs(const MountEntry *m) { |
df6b900a | 943 | const char *entry_path, *inner_path; |
abad72be | 944 | int r; |
abad72be | 945 | |
6c47cd7d LP |
946 | assert(m); |
947 | ||
df6b900a LP |
948 | entry_path = mount_entry_path(m); |
949 | inner_path = m->path_const; | |
950 | ||
2abd4e38 | 951 | /* First, get rid of everything that is below if there is anything. Then, overmount with our new tmpfs */ |
6c47cd7d | 952 | |
abad72be CG |
953 | (void) mkdir_p_label(entry_path, 0755); |
954 | (void) umount_recursive(entry_path, 0); | |
6c47cd7d | 955 | |
21935150 LP |
956 | r = mount_nofollow_verbose(LOG_DEBUG, "tmpfs", entry_path, "tmpfs", m->flags, mount_entry_options(m)); |
957 | if (r < 0) | |
958 | return r; | |
abad72be | 959 | |
df6b900a | 960 | r = label_fix_container(entry_path, inner_path, 0); |
abad72be | 961 | if (r < 0) |
df6b900a | 962 | return log_debug_errno(r, "Failed to fix label of '%s' as '%s': %m", entry_path, inner_path); |
6c47cd7d LP |
963 | |
964 | return 1; | |
965 | } | |
966 | ||
94293d65 LB |
967 | static int mount_run(const MountEntry *m) { |
968 | int r; | |
969 | ||
970 | assert(m); | |
971 | ||
972 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); | |
973 | if (r < 0 && r != -ENOENT) | |
974 | return log_debug_errno(r, "Unable to determine whether /run is already mounted: %m"); | |
975 | if (r > 0) /* make this a NOP if /run is already a mount point */ | |
976 | return 0; | |
977 | ||
978 | return mount_tmpfs(m); | |
979 | } | |
980 | ||
b3d13314 | 981 | static int mount_images(const MountEntry *m) { |
b3d13314 LB |
982 | int r; |
983 | ||
89e62e0b LP |
984 | assert(m); |
985 | ||
4beda316 | 986 | r = verity_dissect_and_mount(mount_entry_source(m), mount_entry_path(m), m->image_options); |
b850a9b2 LB |
987 | if (r == -ENOENT && m->ignore) |
988 | return 0; | |
b3d13314 | 989 | if (r < 0) |
4beda316 | 990 | return log_debug_errno(r, "Failed to mount image %s on %s: %m", mount_entry_source(m), mount_entry_path(m)); |
b3d13314 LB |
991 | |
992 | return 1; | |
993 | } | |
994 | ||
088696fe | 995 | static int follow_symlink( |
d2d6c096 | 996 | const char *root_directory, |
088696fe | 997 | MountEntry *m) { |
d2d6c096 | 998 | |
088696fe | 999 | _cleanup_free_ char *target = NULL; |
8fceda93 LP |
1000 | int r; |
1001 | ||
088696fe LP |
1002 | /* Let's chase symlinks, but only one step at a time. That's because depending where the symlink points we |
1003 | * might need to change the order in which we mount stuff. Hence: let's normalize piecemeal, and do one step at | |
1004 | * a time by specifying CHASE_STEP. This function returns 0 if we resolved one step, and > 0 if we reached the | |
1005 | * end and already have a fully normalized name. */ | |
8fceda93 | 1006 | |
a5648b80 | 1007 | r = chase_symlinks(mount_entry_path(m), root_directory, CHASE_STEP|CHASE_NONEXISTENT, &target, NULL); |
088696fe LP |
1008 | if (r < 0) |
1009 | return log_debug_errno(r, "Failed to chase symlinks '%s': %m", mount_entry_path(m)); | |
1010 | if (r > 0) /* Reached the end, nothing more to resolve */ | |
1011 | return 1; | |
8fceda93 | 1012 | |
baaa35ad ZJS |
1013 | if (m->n_followed >= CHASE_SYMLINKS_MAX) /* put a boundary on things */ |
1014 | return log_debug_errno(SYNTHETIC_ERRNO(ELOOP), | |
1015 | "Symlink loop on '%s'.", | |
1016 | mount_entry_path(m)); | |
8fceda93 | 1017 | |
088696fe | 1018 | log_debug("Followed mount entry path symlink %s → %s.", mount_entry_path(m), target); |
8fceda93 | 1019 | |
088696fe LP |
1020 | free_and_replace(m->path_malloc, target); |
1021 | m->has_prefix = true; | |
8fceda93 | 1022 | |
088696fe LP |
1023 | m->n_followed ++; |
1024 | ||
1025 | return 0; | |
8fceda93 LP |
1026 | } |
1027 | ||
ac0930c8 | 1028 | static int apply_mount( |
8fceda93 | 1029 | const char *root_directory, |
4e399953 LP |
1030 | MountEntry *m, |
1031 | const NamespaceInfo *ns_info) { | |
ac0930c8 | 1032 | |
e5f10caf | 1033 | _cleanup_free_ char *inaccessible = NULL; |
a227a4be | 1034 | bool rbind = true, make = false; |
15ae422b | 1035 | const char *what; |
15ae422b | 1036 | int r; |
15ae422b | 1037 | |
c17ec25e | 1038 | assert(m); |
4e399953 | 1039 | assert(ns_info); |
15ae422b | 1040 | |
34de407a | 1041 | log_debug("Applying namespace mount on %s", mount_entry_path(m)); |
fe3c2583 | 1042 | |
c17ec25e | 1043 | switch (m->mode) { |
15ae422b | 1044 | |
160cfdbe | 1045 | case INACCESSIBLE: { |
e5f10caf AZ |
1046 | _cleanup_free_ char *tmp = NULL; |
1047 | const char *runtime_dir; | |
160cfdbe | 1048 | struct stat target; |
6d313367 LP |
1049 | |
1050 | /* First, get rid of everything that is below if there | |
1051 | * is anything... Then, overmount it with an | |
c4b41707 | 1052 | * inaccessible path. */ |
34de407a | 1053 | (void) umount_recursive(mount_entry_path(m), 0); |
6d313367 | 1054 | |
088696fe LP |
1055 | if (lstat(mount_entry_path(m), &target) < 0) { |
1056 | if (errno == ENOENT && m->ignore) | |
1057 | return 0; | |
1058 | ||
cbc056c8 ZJS |
1059 | return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m", |
1060 | mount_entry_path(m)); | |
088696fe | 1061 | } |
15ae422b | 1062 | |
e5f10caf | 1063 | if (geteuid() == 0) |
48b747fa | 1064 | runtime_dir = "/run"; |
e5f10caf | 1065 | else { |
48b747fa LP |
1066 | if (asprintf(&tmp, "/run/user/" UID_FMT, geteuid()) < 0) |
1067 | return -ENOMEM; | |
e5f10caf AZ |
1068 | |
1069 | runtime_dir = tmp; | |
1070 | } | |
1071 | ||
1072 | r = mode_to_inaccessible_node(runtime_dir, target.st_mode, &inaccessible); | |
1073 | if (r < 0) | |
baaa35ad ZJS |
1074 | return log_debug_errno(SYNTHETIC_ERRNO(ELOOP), |
1075 | "File type not supported for inaccessible mounts. Note that symlinks are not allowed"); | |
e5f10caf | 1076 | what = inaccessible; |
c4b41707 | 1077 | break; |
160cfdbe | 1078 | } |
fe3c2583 | 1079 | |
15ae422b | 1080 | case READONLY: |
15ae422b | 1081 | case READWRITE: |
1e05071d | 1082 | case READWRITE_IMPLICIT: |
ddc155b2 TM |
1083 | case EXEC: |
1084 | case NOEXEC: | |
8fceda93 | 1085 | r = path_is_mount_point(mount_entry_path(m), root_directory, 0); |
088696fe LP |
1086 | if (r == -ENOENT && m->ignore) |
1087 | return 0; | |
d944dc95 | 1088 | if (r < 0) |
cbc056c8 ZJS |
1089 | return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m", |
1090 | mount_entry_path(m)); | |
1091 | if (r > 0) /* Nothing to do here, it is already a mount. We just later toggle the MS_RDONLY | |
ddc155b2 | 1092 | * and MS_NOEXEC bits for the mount point if needed. */ |
6b7c9f8b | 1093 | return 0; |
6b7c9f8b | 1094 | /* This isn't a mount point yet, let's make it one. */ |
34de407a | 1095 | what = mount_entry_path(m); |
6b7c9f8b | 1096 | break; |
15ae422b | 1097 | |
d2d6c096 LP |
1098 | case BIND_MOUNT: |
1099 | rbind = false; | |
d2d6c096 | 1100 | |
4831981d | 1101 | _fallthrough_; |
088696fe LP |
1102 | case BIND_MOUNT_RECURSIVE: { |
1103 | _cleanup_free_ char *chased = NULL; | |
5d997827 | 1104 | |
cbc056c8 ZJS |
1105 | /* Since mount() will always follow symlinks we chase the symlinks on our own first. Note |
1106 | * that bind mount source paths are always relative to the host root, hence we pass NULL as | |
1107 | * root directory to chase_symlinks() here. */ | |
088696fe | 1108 | |
a5648b80 | 1109 | r = chase_symlinks(mount_entry_source(m), NULL, CHASE_TRAIL_SLASH, &chased, NULL); |
088696fe LP |
1110 | if (r == -ENOENT && m->ignore) { |
1111 | log_debug_errno(r, "Path %s does not exist, ignoring.", mount_entry_source(m)); | |
1112 | return 0; | |
1113 | } | |
1114 | if (r < 0) | |
1115 | return log_debug_errno(r, "Failed to follow symlinks on %s: %m", mount_entry_source(m)); | |
1116 | ||
1117 | log_debug("Followed source symlinks %s → %s.", mount_entry_source(m), chased); | |
1118 | ||
1119 | free_and_replace(m->source_malloc, chased); | |
d2d6c096 LP |
1120 | |
1121 | what = mount_entry_source(m); | |
a227a4be | 1122 | make = true; |
d2d6c096 | 1123 | break; |
088696fe | 1124 | } |
d2d6c096 | 1125 | |
6c47cd7d | 1126 | case EMPTY_DIR: |
2abd4e38 YW |
1127 | case TMPFS: |
1128 | return mount_tmpfs(m); | |
6c47cd7d | 1129 | |
ac0930c8 | 1130 | case PRIVATE_TMP: |
56a13a49 | 1131 | case PRIVATE_TMP_READONLY: |
89bd586c | 1132 | what = mount_entry_source(m); |
a227a4be | 1133 | make = true; |
15ae422b | 1134 | break; |
e364ad06 | 1135 | |
d6797c92 | 1136 | case PRIVATE_DEV: |
5d997827 LP |
1137 | return mount_private_dev(m); |
1138 | ||
1139 | case BIND_DEV: | |
1140 | return mount_bind_dev(m); | |
1141 | ||
1142 | case SYSFS: | |
1143 | return mount_sysfs(m); | |
1144 | ||
1145 | case PROCFS: | |
4e399953 | 1146 | return mount_procfs(m, ns_info); |
d6797c92 | 1147 | |
94293d65 LB |
1148 | case RUN: |
1149 | return mount_run(m); | |
1150 | ||
b3d13314 LB |
1151 | case MOUNT_IMAGES: |
1152 | return mount_images(m); | |
1153 | ||
e364ad06 LP |
1154 | default: |
1155 | assert_not_reached("Unknown mode"); | |
15ae422b LP |
1156 | } |
1157 | ||
ac0930c8 | 1158 | assert(what); |
15ae422b | 1159 | |
21935150 LP |
1160 | r = mount_nofollow_verbose(LOG_DEBUG, what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL); |
1161 | if (r < 0) { | |
a227a4be | 1162 | bool try_again = false; |
a227a4be LP |
1163 | |
1164 | if (r == -ENOENT && make) { | |
8bab8029 | 1165 | int q; |
a227a4be | 1166 | |
cbc056c8 ZJS |
1167 | /* Hmm, either the source or the destination are missing. Let's see if we can create |
1168 | the destination, then try again. */ | |
a227a4be | 1169 | |
8bab8029 | 1170 | (void) mkdir_parents(mount_entry_path(m), 0755); |
a227a4be | 1171 | |
8bab8029 LB |
1172 | q = make_mount_point_inode_from_path(what, mount_entry_path(m), 0755); |
1173 | if (q < 0) | |
1174 | log_error_errno(q, "Failed to create destination mount point node '%s': %m", | |
1175 | mount_entry_path(m)); | |
1176 | else | |
1177 | try_again = true; | |
a227a4be LP |
1178 | } |
1179 | ||
21935150 LP |
1180 | if (try_again) |
1181 | r = mount_nofollow_verbose(LOG_DEBUG, what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL); | |
a227a4be | 1182 | if (r < 0) |
5dc60faa | 1183 | return log_error_errno(r, "Failed to mount %s to %s: %m", what, mount_entry_path(m)); |
a227a4be | 1184 | } |
6b7c9f8b | 1185 | |
34de407a | 1186 | log_debug("Successfully mounted %s to %s", what, mount_entry_path(m)); |
6b7c9f8b | 1187 | return 0; |
ac0930c8 | 1188 | } |
15ae422b | 1189 | |
6b000af4 | 1190 | static int make_read_only(const MountEntry *m, char **deny_list, FILE *proc_self_mountinfo) { |
9ce4e4b0 | 1191 | unsigned long new_flags = 0, flags_mask = 0; |
763a260a | 1192 | bool submounts = false; |
6b7c9f8b | 1193 | int r = 0; |
15ae422b | 1194 | |
c17ec25e | 1195 | assert(m); |
ac9de0b3 | 1196 | assert(proc_self_mountinfo); |
ac0930c8 | 1197 | |
9ce4e4b0 LP |
1198 | if (mount_entry_read_only(m) || m->mode == PRIVATE_DEV) { |
1199 | new_flags |= MS_RDONLY; | |
1200 | flags_mask |= MS_RDONLY; | |
1201 | } | |
1202 | ||
1203 | if (m->nosuid) { | |
1204 | new_flags |= MS_NOSUID; | |
1205 | flags_mask |= MS_NOSUID; | |
1206 | } | |
1207 | ||
1208 | if (flags_mask == 0) /* No Change? */ | |
6b7c9f8b LP |
1209 | return 0; |
1210 | ||
9ce4e4b0 LP |
1211 | /* We generally apply these changes recursively, except for /dev, and the cases we know there's |
1212 | * nothing further down. Set /dev readonly, but not submounts like /dev/shm. Also, we only set the | |
1213 | * per-mount read-only flag. We can't set it on the superblock, if we are inside a user namespace | |
1214 | * and running Linux <= 4.17. */ | |
1215 | submounts = | |
1216 | mount_entry_read_only(m) && | |
1217 | !IN_SET(m->mode, EMPTY_DIR, TMPFS); | |
1218 | if (submounts) | |
6b000af4 | 1219 | r = bind_remount_recursive_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, deny_list, proc_self_mountinfo); |
9ce4e4b0 | 1220 | else |
7cce68e1 | 1221 | r = bind_remount_one_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, proc_self_mountinfo); |
9ce4e4b0 | 1222 | |
ddc155b2 | 1223 | /* Note that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked |
867189b5 LP |
1224 | * read-only already stays this way. This improves compatibility with container managers, where we |
1225 | * won't attempt to undo read-only mounts already applied. */ | |
ac0930c8 | 1226 | |
8fceda93 | 1227 | if (r == -ENOENT && m->ignore) |
867189b5 | 1228 | return 0; |
763a260a | 1229 | if (r < 0) |
9ce4e4b0 | 1230 | return log_debug_errno(r, "Failed to re-mount '%s'%s: %m", mount_entry_path(m), |
763a260a | 1231 | submounts ? " and its submounts" : ""); |
763a260a | 1232 | return 0; |
d944dc95 LP |
1233 | } |
1234 | ||
ddc155b2 TM |
1235 | static int make_noexec(const MountEntry *m, char **deny_list, FILE *proc_self_mountinfo) { |
1236 | unsigned long new_flags = 0, flags_mask = 0; | |
1237 | bool submounts = false; | |
1238 | int r = 0; | |
1239 | ||
1240 | assert(m); | |
1241 | assert(proc_self_mountinfo); | |
1242 | ||
1243 | if (mount_entry_noexec(m)) { | |
1244 | new_flags |= MS_NOEXEC; | |
1245 | flags_mask |= MS_NOEXEC; | |
1246 | } else if (mount_entry_exec(m)) { | |
1247 | new_flags &= ~MS_NOEXEC; | |
1248 | flags_mask |= MS_NOEXEC; | |
1249 | } | |
1250 | ||
1251 | if (flags_mask == 0) /* No Change? */ | |
1252 | return 0; | |
1253 | ||
1254 | submounts = !IN_SET(m->mode, EMPTY_DIR, TMPFS); | |
1255 | ||
1256 | if (submounts) | |
1257 | r = bind_remount_recursive_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, deny_list, proc_self_mountinfo); | |
1258 | else | |
1259 | r = bind_remount_one_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, proc_self_mountinfo); | |
1260 | ||
1261 | if (r == -ENOENT && m->ignore) | |
1262 | return 0; | |
1263 | if (r < 0) | |
1264 | return log_debug_errno(r, "Failed to re-mount '%s'%s: %m", mount_entry_path(m), | |
1265 | submounts ? " and its submounts" : ""); | |
1266 | return 0; | |
1267 | } | |
1268 | ||
9b68367b | 1269 | static bool namespace_info_mount_apivfs(const NamespaceInfo *ns_info) { |
5d997827 LP |
1270 | assert(ns_info); |
1271 | ||
9c988f93 DH |
1272 | /* |
1273 | * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=, | |
1274 | * since to protect the API VFS mounts, they need to be around in the | |
9b68367b | 1275 | * first place... |
9c988f93 | 1276 | */ |
5d997827 | 1277 | |
9b68367b YW |
1278 | return ns_info->mount_apivfs || |
1279 | ns_info->protect_control_groups || | |
4e399953 LP |
1280 | ns_info->protect_kernel_tunables || |
1281 | ns_info->protect_proc != PROTECT_PROC_DEFAULT || | |
1282 | ns_info->proc_subset != PROC_SUBSET_ALL; | |
5d997827 LP |
1283 | } |
1284 | ||
da6053d0 | 1285 | static size_t namespace_calculate_mounts( |
bb0ff3fb | 1286 | const NamespaceInfo *ns_info, |
2652c6c1 DH |
1287 | char** read_write_paths, |
1288 | char** read_only_paths, | |
1289 | char** inaccessible_paths, | |
ddc155b2 TM |
1290 | char** exec_paths, |
1291 | char** no_exec_paths, | |
6c47cd7d | 1292 | char** empty_directories, |
da6053d0 LP |
1293 | size_t n_bind_mounts, |
1294 | size_t n_temporary_filesystems, | |
b3d13314 | 1295 | size_t n_mount_images, |
2652c6c1 DH |
1296 | const char* tmp_dir, |
1297 | const char* var_tmp_dir, | |
bbb4e7f3 | 1298 | const char *creds_path, |
5e8deb94 | 1299 | const char* log_namespace, |
3bdc25a4 LP |
1300 | bool setup_propagate, |
1301 | const char* notify_socket) { | |
2652c6c1 | 1302 | |
da6053d0 LP |
1303 | size_t protect_home_cnt; |
1304 | size_t protect_system_cnt = | |
52b3d652 | 1305 | (ns_info->protect_system == PROTECT_SYSTEM_STRICT ? |
f471b2af | 1306 | ELEMENTSOF(protect_system_strict_table) : |
52b3d652 | 1307 | ((ns_info->protect_system == PROTECT_SYSTEM_FULL) ? |
f471b2af | 1308 | ELEMENTSOF(protect_system_full_table) : |
52b3d652 | 1309 | ((ns_info->protect_system == PROTECT_SYSTEM_YES) ? |
f471b2af DH |
1310 | ELEMENTSOF(protect_system_yes_table) : 0))); |
1311 | ||
b6c432ca | 1312 | protect_home_cnt = |
52b3d652 | 1313 | (ns_info->protect_home == PROTECT_HOME_YES ? |
b6c432ca | 1314 | ELEMENTSOF(protect_home_yes_table) : |
52b3d652 | 1315 | ((ns_info->protect_home == PROTECT_HOME_READ_ONLY) ? |
e4da7d8c | 1316 | ELEMENTSOF(protect_home_read_only_table) : |
52b3d652 | 1317 | ((ns_info->protect_home == PROTECT_HOME_TMPFS) ? |
e4da7d8c | 1318 | ELEMENTSOF(protect_home_tmpfs_table) : 0))); |
b6c432ca | 1319 | |
2652c6c1 DH |
1320 | return !!tmp_dir + !!var_tmp_dir + |
1321 | strv_length(read_write_paths) + | |
1322 | strv_length(read_only_paths) + | |
1323 | strv_length(inaccessible_paths) + | |
ddc155b2 TM |
1324 | strv_length(exec_paths) + |
1325 | strv_length(no_exec_paths) + | |
6c47cd7d | 1326 | strv_length(empty_directories) + |
d2d6c096 | 1327 | n_bind_mounts + |
b3d13314 | 1328 | n_mount_images + |
2abd4e38 | 1329 | n_temporary_filesystems + |
c575770b DH |
1330 | ns_info->private_dev + |
1331 | (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) + | |
c575770b | 1332 | (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) + |
94a7b275 KK |
1333 | (ns_info->protect_kernel_logs ? ELEMENTSOF(protect_kernel_logs_table) : 0) + |
1334 | (ns_info->protect_control_groups ? 1 : 0) + | |
5d997827 | 1335 | protect_home_cnt + protect_system_cnt + |
aecd5ac6 | 1336 | (ns_info->protect_hostname ? 2 : 0) + |
91dd5f7c | 1337 | (namespace_info_mount_apivfs(ns_info) ? ELEMENTSOF(apivfs_table) : 0) + |
bbb4e7f3 | 1338 | (creds_path ? 2 : 1) + |
3bdc25a4 LP |
1339 | !!log_namespace + |
1340 | setup_propagate + /* /run/systemd/incoming */ | |
1341 | !!notify_socket; | |
2652c6c1 DH |
1342 | } |
1343 | ||
da6053d0 | 1344 | static void normalize_mounts(const char *root_directory, MountEntry *mounts, size_t *n_mounts) { |
9b68367b | 1345 | assert(root_directory); |
f8b64b57 LP |
1346 | assert(n_mounts); |
1347 | assert(mounts || *n_mounts == 0); | |
1348 | ||
93bab288 | 1349 | typesafe_qsort(mounts, *n_mounts, mount_path_compare); |
f8b64b57 LP |
1350 | |
1351 | drop_duplicates(mounts, n_mounts); | |
1352 | drop_outside_root(root_directory, mounts, n_mounts); | |
1353 | drop_inaccessible(mounts, n_mounts); | |
1354 | drop_nop(mounts, n_mounts); | |
1355 | } | |
1356 | ||
c8c535d5 LP |
1357 | static bool root_read_only( |
1358 | char **read_only_paths, | |
1359 | ProtectSystem protect_system) { | |
1360 | ||
1361 | /* Determine whether the root directory is going to be read-only given the configured settings. */ | |
1362 | ||
1363 | if (protect_system == PROTECT_SYSTEM_STRICT) | |
1364 | return true; | |
1365 | ||
de46b2be | 1366 | if (prefixed_path_strv_contains(read_only_paths, "/")) |
c8c535d5 LP |
1367 | return true; |
1368 | ||
1369 | return false; | |
1370 | } | |
1371 | ||
1372 | static bool home_read_only( | |
1373 | char** read_only_paths, | |
1374 | char** inaccessible_paths, | |
1375 | char** empty_directories, | |
1376 | const BindMount *bind_mounts, | |
1377 | size_t n_bind_mounts, | |
1378 | const TemporaryFileSystem *temporary_filesystems, | |
1379 | size_t n_temporary_filesystems, | |
1380 | ProtectHome protect_home) { | |
1381 | ||
1382 | size_t i; | |
1383 | ||
1384 | /* Determine whether the /home directory is going to be read-only given the configured settings. Yes, | |
1385 | * this is a bit sloppy, since we don't bother checking for cases where / is affected by multiple | |
1386 | * settings. */ | |
1387 | ||
1388 | if (protect_home != PROTECT_HOME_NO) | |
1389 | return true; | |
1390 | ||
de46b2be TM |
1391 | if (prefixed_path_strv_contains(read_only_paths, "/home") || |
1392 | prefixed_path_strv_contains(inaccessible_paths, "/home") || | |
1393 | prefixed_path_strv_contains(empty_directories, "/home")) | |
c8c535d5 LP |
1394 | return true; |
1395 | ||
1396 | for (i = 0; i < n_temporary_filesystems; i++) | |
1397 | if (path_equal(temporary_filesystems[i].path, "/home")) | |
1398 | return true; | |
1399 | ||
1400 | /* If /home is overmounted with some dir from the host it's not writable. */ | |
1401 | for (i = 0; i < n_bind_mounts; i++) | |
1402 | if (path_equal(bind_mounts[i].destination, "/home")) | |
1403 | return true; | |
1404 | ||
1405 | return false; | |
1406 | } | |
1407 | ||
89e62e0b LP |
1408 | static int verity_settings_prepare( |
1409 | VeritySettings *verity, | |
1410 | const char *root_image, | |
1411 | const void *root_hash, | |
1412 | size_t root_hash_size, | |
1413 | const char *root_hash_path, | |
1414 | const void *root_hash_sig, | |
1415 | size_t root_hash_sig_size, | |
1416 | const char *root_hash_sig_path, | |
1417 | const char *verity_data_path) { | |
1418 | ||
1419 | int r; | |
1420 | ||
1421 | assert(verity); | |
1422 | ||
1423 | if (root_hash) { | |
1424 | void *d; | |
1425 | ||
1426 | d = memdup(root_hash, root_hash_size); | |
1427 | if (!d) | |
1428 | return -ENOMEM; | |
1429 | ||
1430 | free_and_replace(verity->root_hash, d); | |
1431 | verity->root_hash_size = root_hash_size; | |
aee36b4e | 1432 | verity->designator = PARTITION_ROOT; |
89e62e0b LP |
1433 | } |
1434 | ||
1435 | if (root_hash_sig) { | |
1436 | void *d; | |
1437 | ||
1438 | d = memdup(root_hash_sig, root_hash_sig_size); | |
1439 | if (!d) | |
1440 | return -ENOMEM; | |
1441 | ||
1442 | free_and_replace(verity->root_hash_sig, d); | |
1443 | verity->root_hash_sig_size = root_hash_sig_size; | |
aee36b4e | 1444 | verity->designator = PARTITION_ROOT; |
89e62e0b LP |
1445 | } |
1446 | ||
1447 | if (verity_data_path) { | |
1448 | r = free_and_strdup(&verity->data_path, verity_data_path); | |
1449 | if (r < 0) | |
1450 | return r; | |
1451 | } | |
1452 | ||
1453 | r = verity_settings_load( | |
1454 | verity, | |
1455 | root_image, | |
1456 | root_hash_path, | |
1457 | root_hash_sig_path); | |
1458 | if (r < 0) | |
1459 | return log_debug_errno(r, "Failed to load root hash: %m"); | |
1460 | ||
1461 | return 0; | |
1462 | } | |
1463 | ||
613b411c | 1464 | int setup_namespace( |
ee818b89 | 1465 | const char* root_directory, |
915e6d16 | 1466 | const char* root_image, |
18d73705 | 1467 | const MountOptions *root_image_options, |
bb0ff3fb | 1468 | const NamespaceInfo *ns_info, |
2a624c36 AP |
1469 | char** read_write_paths, |
1470 | char** read_only_paths, | |
1471 | char** inaccessible_paths, | |
ddc155b2 TM |
1472 | char** exec_paths, |
1473 | char** no_exec_paths, | |
6c47cd7d | 1474 | char** empty_directories, |
d2d6c096 | 1475 | const BindMount *bind_mounts, |
da6053d0 | 1476 | size_t n_bind_mounts, |
2abd4e38 | 1477 | const TemporaryFileSystem *temporary_filesystems, |
da6053d0 | 1478 | size_t n_temporary_filesystems, |
b3d13314 LB |
1479 | const MountImage *mount_images, |
1480 | size_t n_mount_images, | |
a004cb4c LP |
1481 | const char* tmp_dir, |
1482 | const char* var_tmp_dir, | |
bbb4e7f3 | 1483 | const char *creds_path, |
91dd5f7c | 1484 | const char *log_namespace, |
915e6d16 | 1485 | unsigned long mount_flags, |
0389f4fa LB |
1486 | const void *root_hash, |
1487 | size_t root_hash_size, | |
1488 | const char *root_hash_path, | |
d4d55b0d LB |
1489 | const void *root_hash_sig, |
1490 | size_t root_hash_sig_size, | |
1491 | const char *root_hash_sig_path, | |
89e62e0b | 1492 | const char *verity_data_path, |
5e8deb94 LB |
1493 | const char *propagate_dir, |
1494 | const char *incoming_dir, | |
3bdc25a4 | 1495 | const char *notify_socket, |
7cc5ef5f ZJS |
1496 | DissectImageFlags dissect_image_flags, |
1497 | char **error_path) { | |
15ae422b | 1498 | |
915e6d16 | 1499 | _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL; |
78ebe980 | 1500 | _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL; |
915e6d16 | 1501 | _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; |
aee36b4e | 1502 | _cleanup_(verity_settings_done) VeritySettings verity = VERITY_SETTINGS_DEFAULT; |
5f7a690a | 1503 | MountEntry *m = NULL, *mounts = NULL; |
5e8deb94 | 1504 | bool require_prefix = false, setup_propagate = false; |
9b68367b | 1505 | const char *root; |
89e62e0b LP |
1506 | size_t n_mounts; |
1507 | int r; | |
15ae422b | 1508 | |
915e6d16 LP |
1509 | assert(ns_info); |
1510 | ||
5e8deb94 LB |
1511 | if (!isempty(propagate_dir) && !isempty(incoming_dir)) |
1512 | setup_propagate = true; | |
1513 | ||
613b411c | 1514 | if (mount_flags == 0) |
c17ec25e | 1515 | mount_flags = MS_SHARED; |
ac0930c8 | 1516 | |
915e6d16 LP |
1517 | if (root_image) { |
1518 | dissect_image_flags |= DISSECT_IMAGE_REQUIRE_ROOT; | |
1519 | ||
c8c535d5 LP |
1520 | /* Make the whole image read-only if we can determine that we only access it in a read-only fashion. */ |
1521 | if (root_read_only(read_only_paths, | |
52b3d652 | 1522 | ns_info->protect_system) && |
c8c535d5 LP |
1523 | home_read_only(read_only_paths, inaccessible_paths, empty_directories, |
1524 | bind_mounts, n_bind_mounts, temporary_filesystems, n_temporary_filesystems, | |
52b3d652 | 1525 | ns_info->protect_home) && |
c9ef8573 | 1526 | strv_isempty(read_write_paths)) |
915e6d16 LP |
1527 | dissect_image_flags |= DISSECT_IMAGE_READ_ONLY; |
1528 | ||
89e62e0b LP |
1529 | r = verity_settings_prepare( |
1530 | &verity, | |
1531 | root_image, | |
1532 | root_hash, root_hash_size, root_hash_path, | |
1533 | root_hash_sig, root_hash_sig_size, root_hash_sig_path, | |
1534 | verity_data_path); | |
915e6d16 | 1535 | if (r < 0) |
89e62e0b LP |
1536 | return r; |
1537 | ||
1538 | SET_FLAG(dissect_image_flags, DISSECT_IMAGE_NO_PARTITION_TABLE, verity.data_path); | |
915e6d16 | 1539 | |
89e62e0b LP |
1540 | r = loop_device_make_by_path( |
1541 | root_image, | |
1542 | FLAGS_SET(dissect_image_flags, DISSECT_IMAGE_READ_ONLY) ? O_RDONLY : -1 /* < 0 means writable if possible, read-only as fallback */, | |
1543 | FLAGS_SET(dissect_image_flags, DISSECT_IMAGE_NO_PARTITION_TABLE) ? 0 : LO_FLAGS_PARTSCAN, | |
1544 | &loop_device); | |
78ebe980 | 1545 | if (r < 0) |
89e62e0b LP |
1546 | return log_debug_errno(r, "Failed to create loop device for root image: %m"); |
1547 | ||
1548 | r = dissect_image( | |
1549 | loop_device->fd, | |
1550 | &verity, | |
1551 | root_image_options, | |
1552 | dissect_image_flags, | |
1553 | &dissected_image); | |
78ebe980 | 1554 | if (r < 0) |
763a260a | 1555 | return log_debug_errno(r, "Failed to dissect image: %m"); |
78ebe980 | 1556 | |
89e62e0b LP |
1557 | r = dissected_image_decrypt( |
1558 | dissected_image, | |
1559 | NULL, | |
1560 | &verity, | |
1561 | dissect_image_flags, | |
1562 | &decrypted_image); | |
915e6d16 | 1563 | if (r < 0) |
763a260a | 1564 | return log_debug_errno(r, "Failed to decrypt dissected image: %m"); |
915e6d16 LP |
1565 | } |
1566 | ||
e908468b LP |
1567 | if (root_directory) |
1568 | root = root_directory; | |
0722b359 | 1569 | else { |
77f16dbd DDM |
1570 | /* /run/systemd should have been created by PID 1 early on already, but in some cases, like |
1571 | * when running tests (test-execute), it might not have been created yet so let's make sure | |
1572 | * we create it if it doesn't already exist. */ | |
1573 | (void) mkdir_p_label("/run/systemd", 0755); | |
1574 | ||
0722b359 JS |
1575 | /* Always create the mount namespace in a temporary directory, instead of operating |
1576 | * directly in the root. The temporary directory prevents any mounts from being | |
1577 | * potentially obscured my other mounts we already applied. | |
1578 | * We use the same mount point for all images, which is safe, since they all live | |
1579 | * in their own namespaces after all, and hence won't see each other. */ | |
e908468b LP |
1580 | |
1581 | root = "/run/systemd/unit-root"; | |
1582 | (void) mkdir_label(root, 0700); | |
d18aff04 | 1583 | require_prefix = true; |
0722b359 | 1584 | } |
e908468b | 1585 | |
cfbeb4ef LP |
1586 | n_mounts = namespace_calculate_mounts( |
1587 | ns_info, | |
1588 | read_write_paths, | |
1589 | read_only_paths, | |
1590 | inaccessible_paths, | |
ddc155b2 TM |
1591 | exec_paths, |
1592 | no_exec_paths, | |
6c47cd7d | 1593 | empty_directories, |
f5c52a77 | 1594 | n_bind_mounts, |
2abd4e38 | 1595 | n_temporary_filesystems, |
b3d13314 | 1596 | n_mount_images, |
cfbeb4ef | 1597 | tmp_dir, var_tmp_dir, |
bbb4e7f3 | 1598 | creds_path, |
5e8deb94 | 1599 | log_namespace, |
3bdc25a4 LP |
1600 | setup_propagate, |
1601 | notify_socket); | |
613b411c | 1602 | |
f0a4feb0 | 1603 | if (n_mounts > 0) { |
5f7a690a LP |
1604 | m = mounts = new0(MountEntry, n_mounts); |
1605 | if (!mounts) | |
1606 | return -ENOMEM; | |
1607 | ||
d18aff04 | 1608 | r = append_access_mounts(&m, read_write_paths, READWRITE, require_prefix); |
613b411c | 1609 | if (r < 0) |
f0a4feb0 | 1610 | goto finish; |
613b411c | 1611 | |
d18aff04 | 1612 | r = append_access_mounts(&m, read_only_paths, READONLY, require_prefix); |
613b411c | 1613 | if (r < 0) |
f0a4feb0 | 1614 | goto finish; |
613b411c | 1615 | |
d18aff04 | 1616 | r = append_access_mounts(&m, inaccessible_paths, INACCESSIBLE, require_prefix); |
613b411c | 1617 | if (r < 0) |
f0a4feb0 | 1618 | goto finish; |
7ff7394d | 1619 | |
ddc155b2 TM |
1620 | r = append_access_mounts(&m, exec_paths, EXEC, require_prefix); |
1621 | if (r < 0) | |
1622 | goto finish; | |
1623 | ||
1624 | r = append_access_mounts(&m, no_exec_paths, NOEXEC, require_prefix); | |
1625 | if (r < 0) | |
1626 | goto finish; | |
1627 | ||
6c47cd7d LP |
1628 | r = append_empty_dir_mounts(&m, empty_directories); |
1629 | if (r < 0) | |
1630 | goto finish; | |
1631 | ||
d2d6c096 LP |
1632 | r = append_bind_mounts(&m, bind_mounts, n_bind_mounts); |
1633 | if (r < 0) | |
1634 | goto finish; | |
1635 | ||
2abd4e38 YW |
1636 | r = append_tmpfs_mounts(&m, temporary_filesystems, n_temporary_filesystems); |
1637 | if (r < 0) | |
1638 | goto finish; | |
1639 | ||
613b411c | 1640 | if (tmp_dir) { |
56a13a49 ZJS |
1641 | bool ro = streq(tmp_dir, RUN_SYSTEMD_EMPTY); |
1642 | ||
34de407a | 1643 | *(m++) = (MountEntry) { |
5327c910 | 1644 | .path_const = "/tmp", |
56a13a49 | 1645 | .mode = ro ? PRIVATE_TMP_READONLY : PRIVATE_TMP, |
89bd586c | 1646 | .source_const = tmp_dir, |
5327c910 | 1647 | }; |
613b411c | 1648 | } |
7ff7394d | 1649 | |
613b411c | 1650 | if (var_tmp_dir) { |
56a13a49 ZJS |
1651 | bool ro = streq(var_tmp_dir, RUN_SYSTEMD_EMPTY); |
1652 | ||
34de407a | 1653 | *(m++) = (MountEntry) { |
5327c910 | 1654 | .path_const = "/var/tmp", |
56a13a49 | 1655 | .mode = ro ? PRIVATE_TMP_READONLY : PRIVATE_TMP, |
89bd586c | 1656 | .source_const = var_tmp_dir, |
5327c910 | 1657 | }; |
7ff7394d | 1658 | } |
ac0930c8 | 1659 | |
b3d13314 LB |
1660 | r = append_mount_images(&m, mount_images, n_mount_images); |
1661 | if (r < 0) | |
1662 | goto finish; | |
1663 | ||
d46b79bb | 1664 | if (ns_info->private_dev) |
34de407a | 1665 | *(m++) = (MountEntry) { |
5327c910 LP |
1666 | .path_const = "/dev", |
1667 | .mode = PRIVATE_DEV, | |
9ce4e4b0 | 1668 | .flags = DEV_MOUNT_OPTIONS, |
5327c910 | 1669 | }; |
7f112f50 | 1670 | |
c575770b | 1671 | if (ns_info->protect_kernel_tunables) { |
cbc056c8 ZJS |
1672 | r = append_static_mounts(&m, |
1673 | protect_kernel_tunables_table, | |
1674 | ELEMENTSOF(protect_kernel_tunables_table), | |
1675 | ns_info->ignore_protect_paths); | |
c575770b | 1676 | if (r < 0) |
f0a4feb0 | 1677 | goto finish; |
c575770b DH |
1678 | } |
1679 | ||
1680 | if (ns_info->protect_kernel_modules) { | |
cbc056c8 ZJS |
1681 | r = append_static_mounts(&m, |
1682 | protect_kernel_modules_table, | |
1683 | ELEMENTSOF(protect_kernel_modules_table), | |
1684 | ns_info->ignore_protect_paths); | |
c575770b | 1685 | if (r < 0) |
f0a4feb0 | 1686 | goto finish; |
c575770b | 1687 | } |
59eeb84b | 1688 | |
94a7b275 | 1689 | if (ns_info->protect_kernel_logs) { |
cbc056c8 ZJS |
1690 | r = append_static_mounts(&m, |
1691 | protect_kernel_logs_table, | |
1692 | ELEMENTSOF(protect_kernel_logs_table), | |
1693 | ns_info->ignore_protect_paths); | |
94a7b275 KK |
1694 | if (r < 0) |
1695 | goto finish; | |
1696 | } | |
1697 | ||
d46b79bb | 1698 | if (ns_info->protect_control_groups) |
34de407a | 1699 | *(m++) = (MountEntry) { |
5327c910 LP |
1700 | .path_const = "/sys/fs/cgroup", |
1701 | .mode = READONLY, | |
1702 | }; | |
59eeb84b | 1703 | |
52b3d652 | 1704 | r = append_protect_home(&m, ns_info->protect_home, ns_info->ignore_protect_paths); |
b6c432ca | 1705 | if (r < 0) |
f0a4feb0 | 1706 | goto finish; |
417116f2 | 1707 | |
52b3d652 | 1708 | r = append_protect_system(&m, ns_info->protect_system, false); |
f471b2af | 1709 | if (r < 0) |
f0a4feb0 | 1710 | goto finish; |
417116f2 | 1711 | |
9b68367b | 1712 | if (namespace_info_mount_apivfs(ns_info)) { |
cbc056c8 ZJS |
1713 | r = append_static_mounts(&m, |
1714 | apivfs_table, | |
1715 | ELEMENTSOF(apivfs_table), | |
1716 | ns_info->ignore_protect_paths); | |
5d997827 LP |
1717 | if (r < 0) |
1718 | goto finish; | |
1719 | } | |
1720 | ||
aecd5ac6 TM |
1721 | if (ns_info->protect_hostname) { |
1722 | *(m++) = (MountEntry) { | |
1723 | .path_const = "/proc/sys/kernel/hostname", | |
1724 | .mode = READONLY, | |
1725 | }; | |
1726 | *(m++) = (MountEntry) { | |
1727 | .path_const = "/proc/sys/kernel/domainname", | |
1728 | .mode = READONLY, | |
1729 | }; | |
1730 | } | |
1731 | ||
bbb4e7f3 LP |
1732 | if (creds_path) { |
1733 | /* If our service has a credentials store configured, then bind that one in, but hide | |
1734 | * everything else. */ | |
1735 | ||
1736 | *(m++) = (MountEntry) { | |
1737 | .path_const = "/run/credentials", | |
1738 | .mode = TMPFS, | |
1739 | .read_only = true, | |
1740 | .options_const = "mode=0755" TMPFS_LIMITS_EMPTY_OR_ALMOST, | |
1741 | .flags = MS_NODEV|MS_STRICTATIME|MS_NOSUID|MS_NOEXEC, | |
1742 | }; | |
1743 | ||
1744 | *(m++) = (MountEntry) { | |
1745 | .path_const = creds_path, | |
1746 | .mode = BIND_MOUNT, | |
1747 | .read_only = true, | |
1748 | .source_const = creds_path, | |
1749 | }; | |
1750 | } else { | |
1751 | /* If our service has no credentials store configured, then make the whole | |
1752 | * credentials tree inaccessible wholesale. */ | |
1753 | ||
1754 | *(m++) = (MountEntry) { | |
1755 | .path_const = "/run/credentials", | |
1756 | .mode = INACCESSIBLE, | |
1757 | .ignore = true, | |
1758 | }; | |
1759 | } | |
1760 | ||
91dd5f7c LP |
1761 | if (log_namespace) { |
1762 | _cleanup_free_ char *q; | |
1763 | ||
1764 | q = strjoin("/run/systemd/journal.", log_namespace); | |
1765 | if (!q) { | |
1766 | r = -ENOMEM; | |
1767 | goto finish; | |
1768 | } | |
1769 | ||
1770 | *(m++) = (MountEntry) { | |
1771 | .path_const = "/run/systemd/journal", | |
1772 | .mode = BIND_MOUNT_RECURSIVE, | |
1773 | .read_only = true, | |
1774 | .source_malloc = TAKE_PTR(q), | |
1775 | }; | |
1776 | } | |
1777 | ||
5e8deb94 LB |
1778 | /* Will be used to add bind mounts at runtime */ |
1779 | if (setup_propagate) | |
1780 | *(m++) = (MountEntry) { | |
1781 | .source_const = propagate_dir, | |
1782 | .path_const = incoming_dir, | |
1783 | .mode = BIND_MOUNT, | |
1784 | .read_only = true, | |
1785 | }; | |
1786 | ||
3bdc25a4 LP |
1787 | if (notify_socket) |
1788 | *(m++) = (MountEntry) { | |
1789 | .path_const = notify_socket, | |
1790 | .source_const = notify_socket, | |
1791 | .mode = BIND_MOUNT, | |
1792 | .read_only = true, | |
1793 | }; | |
1794 | ||
f0a4feb0 | 1795 | assert(mounts + n_mounts == m); |
ac0930c8 | 1796 | |
5327c910 | 1797 | /* Prepend the root directory where that's necessary */ |
e908468b | 1798 | r = prefix_where_needed(mounts, n_mounts, root); |
5327c910 LP |
1799 | if (r < 0) |
1800 | goto finish; | |
1801 | ||
839f1877 | 1802 | normalize_mounts(root, mounts, &n_mounts); |
15ae422b LP |
1803 | } |
1804 | ||
1beab8b0 LP |
1805 | /* All above is just preparation, figuring out what to do. Let's now actually start doing something. */ |
1806 | ||
d944dc95 | 1807 | if (unshare(CLONE_NEWNS) < 0) { |
763a260a | 1808 | r = log_debug_errno(errno, "Failed to unshare the mount namespace: %m"); |
1beab8b0 | 1809 | if (IN_SET(r, -EACCES, -EPERM, -EOPNOTSUPP, -ENOSYS)) |
cbc056c8 ZJS |
1810 | /* If the kernel doesn't support namespaces, or when there's a MAC or seccomp filter |
1811 | * in place that doesn't allow us to create namespaces (or a missing cap), then | |
1812 | * propagate a recognizable error back, which the caller can use to detect this case | |
1813 | * (and only this) and optionally continue without namespacing applied. */ | |
1beab8b0 LP |
1814 | r = -ENOANO; |
1815 | ||
d944dc95 LP |
1816 | goto finish; |
1817 | } | |
1e4e94c8 | 1818 | |
5e8deb94 LB |
1819 | /* Create the source directory to allow runtime propagation of mounts */ |
1820 | if (setup_propagate) | |
1821 | (void) mkdir_p(propagate_dir, 0600); | |
1822 | ||
9b68367b YW |
1823 | /* Remount / as SLAVE so that nothing now mounted in the namespace |
1824 | * shows up in the parent */ | |
1825 | if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { | |
763a260a | 1826 | r = log_debug_errno(errno, "Failed to remount '/' as SLAVE: %m"); |
9b68367b | 1827 | goto finish; |
ee818b89 AC |
1828 | } |
1829 | ||
915e6d16 | 1830 | if (root_image) { |
e908468b | 1831 | /* A root image is specified, mount it to the right place */ |
2d3a5a73 | 1832 | r = dissected_image_mount(dissected_image, root, UID_INVALID, dissect_image_flags); |
763a260a YW |
1833 | if (r < 0) { |
1834 | log_debug_errno(r, "Failed to mount root image: %m"); | |
915e6d16 | 1835 | goto finish; |
763a260a | 1836 | } |
915e6d16 | 1837 | |
07ce7407 TM |
1838 | if (decrypted_image) { |
1839 | r = decrypted_image_relinquish(decrypted_image); | |
763a260a YW |
1840 | if (r < 0) { |
1841 | log_debug_errno(r, "Failed to relinquish decrypted image: %m"); | |
07ce7407 | 1842 | goto finish; |
763a260a | 1843 | } |
07ce7407 | 1844 | } |
78ebe980 | 1845 | |
915e6d16 LP |
1846 | loop_device_relinquish(loop_device); |
1847 | ||
1848 | } else if (root_directory) { | |
1849 | ||
e908468b LP |
1850 | /* A root directory is specified. Turn its directory into bind mount, if it isn't one yet. */ |
1851 | r = path_is_mount_point(root, NULL, AT_SYMLINK_FOLLOW); | |
763a260a YW |
1852 | if (r < 0) { |
1853 | log_debug_errno(r, "Failed to detect that %s is a mount point or not: %m", root); | |
d944dc95 | 1854 | goto finish; |
763a260a | 1855 | } |
8f1ad200 | 1856 | if (r == 0) { |
21935150 LP |
1857 | r = mount_nofollow_verbose(LOG_DEBUG, root, root, NULL, MS_BIND|MS_REC, NULL); |
1858 | if (r < 0) | |
8f1ad200 | 1859 | goto finish; |
d944dc95 | 1860 | } |
e908468b | 1861 | |
9b68367b | 1862 | } else { |
e908468b | 1863 | /* Let's mount the main root directory to the root directory to use */ |
21935150 LP |
1864 | r = mount_nofollow_verbose(LOG_DEBUG, "/", root, NULL, MS_BIND|MS_REC, NULL); |
1865 | if (r < 0) | |
e908468b | 1866 | goto finish; |
ee818b89 | 1867 | } |
c2c13f2d | 1868 | |
4e0c20de LP |
1869 | /* Try to set up the new root directory before mounting anything else there. */ |
1870 | if (root_image || root_directory) | |
1871 | (void) base_filesystem_create(root, UID_INVALID, GID_INVALID); | |
1872 | ||
f0a4feb0 | 1873 | if (n_mounts > 0) { |
ac9de0b3 | 1874 | _cleanup_fclose_ FILE *proc_self_mountinfo = NULL; |
6b000af4 | 1875 | _cleanup_free_ char **deny_list = NULL; |
da6053d0 | 1876 | size_t j; |
6b7c9f8b | 1877 | |
cbc056c8 ZJS |
1878 | /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of |
1879 | * /proc. For example, this is the case with the option: 'InaccessiblePaths=/proc'. */ | |
ac9de0b3 TR |
1880 | proc_self_mountinfo = fopen("/proc/self/mountinfo", "re"); |
1881 | if (!proc_self_mountinfo) { | |
763a260a | 1882 | r = log_debug_errno(errno, "Failed to open /proc/self/mountinfo: %m"); |
7cc5ef5f ZJS |
1883 | if (error_path) |
1884 | *error_path = strdup("/proc/self/mountinfo"); | |
ac9de0b3 TR |
1885 | goto finish; |
1886 | } | |
1887 | ||
088696fe LP |
1888 | /* First round, establish all mounts we need */ |
1889 | for (;;) { | |
1890 | bool again = false; | |
1891 | ||
1892 | for (m = mounts; m < mounts + n_mounts; ++m) { | |
1893 | ||
1894 | if (m->applied) | |
1895 | continue; | |
1896 | ||
1897 | r = follow_symlink(root, m); | |
7cc5ef5f ZJS |
1898 | if (r < 0) { |
1899 | if (error_path && mount_entry_path(m)) | |
1900 | *error_path = strdup(mount_entry_path(m)); | |
088696fe | 1901 | goto finish; |
7cc5ef5f | 1902 | } |
088696fe | 1903 | if (r == 0) { |
cbc056c8 ZJS |
1904 | /* We hit a symlinked mount point. The entry got rewritten and might |
1905 | * point to a very different place now. Let's normalize the changed | |
1906 | * list, and start from the beginning. After all to mount the entry | |
1907 | * at the new location we might need some other mounts first */ | |
088696fe LP |
1908 | again = true; |
1909 | break; | |
1910 | } | |
1911 | ||
4e399953 | 1912 | r = apply_mount(root, m, ns_info); |
7cc5ef5f ZJS |
1913 | if (r < 0) { |
1914 | if (error_path && mount_entry_path(m)) | |
1915 | *error_path = strdup(mount_entry_path(m)); | |
088696fe | 1916 | goto finish; |
7cc5ef5f | 1917 | } |
088696fe LP |
1918 | |
1919 | m->applied = true; | |
1920 | } | |
1921 | ||
1922 | if (!again) | |
1923 | break; | |
1924 | ||
839f1877 | 1925 | normalize_mounts(root, mounts, &n_mounts); |
c2c13f2d | 1926 | } |
15ae422b | 1927 | |
6b000af4 LP |
1928 | /* Create a deny list we can pass to bind_mount_recursive() */ |
1929 | deny_list = new(char*, n_mounts+1); | |
1930 | if (!deny_list) { | |
5f7a690a LP |
1931 | r = -ENOMEM; |
1932 | goto finish; | |
1933 | } | |
f0a4feb0 | 1934 | for (j = 0; j < n_mounts; j++) |
6b000af4 LP |
1935 | deny_list[j] = (char*) mount_entry_path(mounts+j); |
1936 | deny_list[j] = NULL; | |
6b7c9f8b LP |
1937 | |
1938 | /* Second round, flip the ro bits if necessary. */ | |
f0a4feb0 | 1939 | for (m = mounts; m < mounts + n_mounts; ++m) { |
6b000af4 | 1940 | r = make_read_only(m, deny_list, proc_self_mountinfo); |
7cc5ef5f ZJS |
1941 | if (r < 0) { |
1942 | if (error_path && mount_entry_path(m)) | |
1943 | *error_path = strdup(mount_entry_path(m)); | |
d944dc95 | 1944 | goto finish; |
7cc5ef5f | 1945 | } |
c2c13f2d | 1946 | } |
ddc155b2 TM |
1947 | |
1948 | /* Third round, flip the noexec bits with a simplified deny list. */ | |
1949 | for (m = mounts, j = 0; m < mounts + n_mounts; ++m) | |
1950 | if (IN_SET(m->mode, EXEC, NOEXEC)) | |
1951 | deny_list[j++] = (char*) mount_entry_path(m); | |
1952 | deny_list[j] = NULL; | |
1953 | ||
1954 | for (m = mounts; m < mounts + n_mounts; ++m) { | |
1955 | r = make_noexec(m, deny_list, proc_self_mountinfo); | |
1956 | if (r < 0) { | |
1957 | if (error_path && mount_entry_path(m)) | |
1958 | *error_path = strdup(mount_entry_path(m)); | |
1959 | goto finish; | |
1960 | } | |
1961 | } | |
15ae422b LP |
1962 | } |
1963 | ||
9b68367b YW |
1964 | /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */ |
1965 | r = mount_move_root(root); | |
763a260a YW |
1966 | if (r < 0) { |
1967 | log_debug_errno(r, "Failed to mount root with MS_MOVE: %m"); | |
9b68367b | 1968 | goto finish; |
763a260a | 1969 | } |
ee818b89 | 1970 | |
55fe7432 | 1971 | /* Remount / as the desired mode. Note that this will not |
c2c13f2d LP |
1972 | * reestablish propagation from our side to the host, since |
1973 | * what's disconnected is disconnected. */ | |
d944dc95 | 1974 | if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) { |
763a260a | 1975 | r = log_debug_errno(errno, "Failed to remount '/' with desired mount flags: %m"); |
d944dc95 LP |
1976 | goto finish; |
1977 | } | |
15ae422b | 1978 | |
5e8deb94 LB |
1979 | /* bind_mount_in_namespace() will MS_MOVE into that directory, and that's only |
1980 | * supported for non-shared mounts. This needs to happen after remounting / or it will fail. */ | |
1981 | if (setup_propagate) { | |
1982 | r = mount(NULL, incoming_dir, NULL, MS_SLAVE, NULL); | |
1983 | if (r < 0) { | |
1984 | log_error_errno(r, "Failed to remount %s with MS_SLAVE: %m", incoming_dir); | |
1985 | goto finish; | |
1986 | } | |
1987 | } | |
1988 | ||
d944dc95 | 1989 | r = 0; |
15ae422b | 1990 | |
d944dc95 | 1991 | finish: |
0cd41757 LP |
1992 | if (n_mounts > 0) |
1993 | for (m = mounts; m < mounts + n_mounts; m++) | |
1994 | mount_entry_done(m); | |
613b411c | 1995 | |
5f7a690a LP |
1996 | free(mounts); |
1997 | ||
613b411c LP |
1998 | return r; |
1999 | } | |
2000 | ||
da6053d0 | 2001 | void bind_mount_free_many(BindMount *b, size_t n) { |
d2d6c096 LP |
2002 | assert(b || n == 0); |
2003 | ||
fe96c0f8 | 2004 | for (size_t i = 0; i < n; i++) { |
d2d6c096 LP |
2005 | free(b[i].source); |
2006 | free(b[i].destination); | |
2007 | } | |
2008 | ||
2009 | free(b); | |
2010 | } | |
2011 | ||
da6053d0 | 2012 | int bind_mount_add(BindMount **b, size_t *n, const BindMount *item) { |
d2d6c096 LP |
2013 | _cleanup_free_ char *s = NULL, *d = NULL; |
2014 | BindMount *c; | |
2015 | ||
2016 | assert(b); | |
2017 | assert(n); | |
2018 | assert(item); | |
2019 | ||
2020 | s = strdup(item->source); | |
2021 | if (!s) | |
2022 | return -ENOMEM; | |
2023 | ||
2024 | d = strdup(item->destination); | |
2025 | if (!d) | |
2026 | return -ENOMEM; | |
2027 | ||
aa484f35 | 2028 | c = reallocarray(*b, *n + 1, sizeof(BindMount)); |
d2d6c096 LP |
2029 | if (!c) |
2030 | return -ENOMEM; | |
2031 | ||
2032 | *b = c; | |
2033 | ||
2034 | c[(*n) ++] = (BindMount) { | |
1cc6c93a YW |
2035 | .source = TAKE_PTR(s), |
2036 | .destination = TAKE_PTR(d), | |
d2d6c096 | 2037 | .read_only = item->read_only, |
9ce4e4b0 | 2038 | .nosuid = item->nosuid, |
d2d6c096 LP |
2039 | .recursive = item->recursive, |
2040 | .ignore_enoent = item->ignore_enoent, | |
2041 | }; | |
2042 | ||
d2d6c096 LP |
2043 | return 0; |
2044 | } | |
2045 | ||
b3d13314 | 2046 | MountImage* mount_image_free_many(MountImage *m, size_t *n) { |
b3d13314 LB |
2047 | assert(n); |
2048 | assert(m || *n == 0); | |
2049 | ||
fe96c0f8 | 2050 | for (size_t i = 0; i < *n; i++) { |
b3d13314 LB |
2051 | free(m[i].source); |
2052 | free(m[i].destination); | |
427353f6 | 2053 | mount_options_free_all(m[i].mount_options); |
b3d13314 LB |
2054 | } |
2055 | ||
2056 | free(m); | |
2057 | *n = 0; | |
2058 | return NULL; | |
2059 | } | |
2060 | ||
2061 | int mount_image_add(MountImage **m, size_t *n, const MountImage *item) { | |
2062 | _cleanup_free_ char *s = NULL, *d = NULL; | |
427353f6 LB |
2063 | _cleanup_(mount_options_free_allp) MountOptions *options = NULL; |
2064 | MountOptions *i; | |
b3d13314 LB |
2065 | MountImage *c; |
2066 | ||
2067 | assert(m); | |
2068 | assert(n); | |
2069 | assert(item); | |
2070 | ||
2071 | s = strdup(item->source); | |
2072 | if (!s) | |
2073 | return -ENOMEM; | |
2074 | ||
2075 | d = strdup(item->destination); | |
2076 | if (!d) | |
2077 | return -ENOMEM; | |
2078 | ||
427353f6 LB |
2079 | LIST_FOREACH(mount_options, i, item->mount_options) { |
2080 | _cleanup_(mount_options_free_allp) MountOptions *o; | |
2081 | ||
2082 | o = new(MountOptions, 1); | |
2083 | if (!o) | |
2084 | return -ENOMEM; | |
2085 | ||
2086 | *o = (MountOptions) { | |
2087 | .partition_designator = i->partition_designator, | |
2088 | .options = strdup(i->options), | |
2089 | }; | |
2090 | if (!o->options) | |
2091 | return -ENOMEM; | |
2092 | ||
2093 | LIST_APPEND(mount_options, options, TAKE_PTR(o)); | |
2094 | } | |
2095 | ||
b3d13314 LB |
2096 | c = reallocarray(*m, *n + 1, sizeof(MountImage)); |
2097 | if (!c) | |
2098 | return -ENOMEM; | |
2099 | ||
2100 | *m = c; | |
2101 | ||
2102 | c[(*n) ++] = (MountImage) { | |
2103 | .source = TAKE_PTR(s), | |
2104 | .destination = TAKE_PTR(d), | |
427353f6 | 2105 | .mount_options = TAKE_PTR(options), |
b3d13314 LB |
2106 | .ignore_enoent = item->ignore_enoent, |
2107 | }; | |
2108 | ||
2109 | return 0; | |
2110 | } | |
2111 | ||
da6053d0 | 2112 | void temporary_filesystem_free_many(TemporaryFileSystem *t, size_t n) { |
2abd4e38 YW |
2113 | assert(t || n == 0); |
2114 | ||
fe96c0f8 | 2115 | for (size_t i = 0; i < n; i++) { |
2abd4e38 YW |
2116 | free(t[i].path); |
2117 | free(t[i].options); | |
2118 | } | |
2119 | ||
2120 | free(t); | |
2121 | } | |
2122 | ||
2123 | int temporary_filesystem_add( | |
2124 | TemporaryFileSystem **t, | |
da6053d0 | 2125 | size_t *n, |
2abd4e38 YW |
2126 | const char *path, |
2127 | const char *options) { | |
2128 | ||
2129 | _cleanup_free_ char *p = NULL, *o = NULL; | |
2130 | TemporaryFileSystem *c; | |
2131 | ||
2132 | assert(t); | |
2133 | assert(n); | |
2134 | assert(path); | |
2135 | ||
2136 | p = strdup(path); | |
2137 | if (!p) | |
2138 | return -ENOMEM; | |
2139 | ||
2140 | if (!isempty(options)) { | |
2141 | o = strdup(options); | |
2142 | if (!o) | |
2143 | return -ENOMEM; | |
2144 | } | |
2145 | ||
aa484f35 | 2146 | c = reallocarray(*t, *n + 1, sizeof(TemporaryFileSystem)); |
2abd4e38 YW |
2147 | if (!c) |
2148 | return -ENOMEM; | |
2149 | ||
2150 | *t = c; | |
2151 | ||
2152 | c[(*n) ++] = (TemporaryFileSystem) { | |
1cc6c93a YW |
2153 | .path = TAKE_PTR(p), |
2154 | .options = TAKE_PTR(o), | |
2abd4e38 YW |
2155 | }; |
2156 | ||
2abd4e38 YW |
2157 | return 0; |
2158 | } | |
2159 | ||
a652f050 JR |
2160 | static int make_tmp_prefix(const char *prefix) { |
2161 | _cleanup_free_ char *t = NULL; | |
2162 | int r; | |
2163 | ||
2164 | /* Don't do anything unless we know the dir is actually missing */ | |
2165 | r = access(prefix, F_OK); | |
2166 | if (r >= 0) | |
2167 | return 0; | |
2168 | if (errno != ENOENT) | |
2169 | return -errno; | |
2170 | ||
2171 | r = mkdir_parents(prefix, 0755); | |
2172 | if (r < 0) | |
2173 | return r; | |
2174 | ||
2175 | r = tempfn_random(prefix, NULL, &t); | |
2176 | if (r < 0) | |
2177 | return r; | |
2178 | ||
2179 | if (mkdir(t, 0777) < 0) | |
2180 | return -errno; | |
2181 | ||
2182 | if (chmod(t, 01777) < 0) { | |
2183 | r = -errno; | |
2184 | (void) rmdir(t); | |
2185 | return r; | |
2186 | } | |
2187 | ||
2188 | if (rename(t, prefix) < 0) { | |
2189 | r = -errno; | |
2190 | (void) rmdir(t); | |
2191 | return r == -EEXIST ? 0 : r; /* it's fine if someone else created the dir by now */ | |
2192 | } | |
2193 | ||
2194 | return 0; | |
2195 | ||
2196 | } | |
2197 | ||
56a13a49 | 2198 | static int setup_one_tmp_dir(const char *id, const char *prefix, char **path, char **tmp_path) { |
613b411c | 2199 | _cleanup_free_ char *x = NULL; |
19cd4e19 | 2200 | _cleanup_free_ char *y = NULL; |
6b46ea73 LP |
2201 | char bid[SD_ID128_STRING_MAX]; |
2202 | sd_id128_t boot_id; | |
56a13a49 | 2203 | bool rw = true; |
6b46ea73 | 2204 | int r; |
613b411c LP |
2205 | |
2206 | assert(id); | |
2207 | assert(prefix); | |
2208 | assert(path); | |
2209 | ||
6b46ea73 LP |
2210 | /* We include the boot id in the directory so that after a |
2211 | * reboot we can easily identify obsolete directories. */ | |
2212 | ||
2213 | r = sd_id128_get_boot(&boot_id); | |
2214 | if (r < 0) | |
2215 | return r; | |
2216 | ||
605405c6 | 2217 | x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX"); |
613b411c LP |
2218 | if (!x) |
2219 | return -ENOMEM; | |
2220 | ||
a652f050 JR |
2221 | r = make_tmp_prefix(prefix); |
2222 | if (r < 0) | |
2223 | return r; | |
2224 | ||
613b411c | 2225 | RUN_WITH_UMASK(0077) |
56a13a49 ZJS |
2226 | if (!mkdtemp(x)) { |
2227 | if (errno == EROFS || ERRNO_IS_DISK_SPACE(errno)) | |
2228 | rw = false; | |
2229 | else | |
2230 | return -errno; | |
2231 | } | |
613b411c | 2232 | |
56a13a49 | 2233 | if (rw) { |
19cd4e19 | 2234 | y = strjoin(x, "/tmp"); |
2235 | if (!y) | |
2236 | return -ENOMEM; | |
2237 | ||
2238 | RUN_WITH_UMASK(0000) { | |
2239 | if (mkdir(y, 0777 | S_ISVTX) < 0) | |
2240 | return -errno; | |
2241 | } | |
2242 | ||
2243 | r = label_fix_container(y, prefix, 0); | |
56a13a49 ZJS |
2244 | if (r < 0) |
2245 | return r; | |
19cd4e19 | 2246 | |
2247 | if (tmp_path) | |
2248 | *tmp_path = TAKE_PTR(y); | |
56a13a49 ZJS |
2249 | } else { |
2250 | /* Trouble: we failed to create the directory. Instead of failing, let's simulate /tmp being | |
2251 | * read-only. This way the service will get the EROFS result as if it was writing to the real | |
2252 | * file system. */ | |
2253 | r = mkdir_p(RUN_SYSTEMD_EMPTY, 0500); | |
2254 | if (r < 0) | |
2255 | return r; | |
613b411c | 2256 | |
3f181262 LP |
2257 | r = free_and_strdup(&x, RUN_SYSTEMD_EMPTY); |
2258 | if (r < 0) | |
2259 | return r; | |
c17ec25e | 2260 | } |
15ae422b | 2261 | |
1cc6c93a | 2262 | *path = TAKE_PTR(x); |
613b411c LP |
2263 | return 0; |
2264 | } | |
2265 | ||
2266 | int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) { | |
56a13a49 ZJS |
2267 | _cleanup_(namespace_cleanup_tmpdirp) char *a = NULL; |
2268 | _cleanup_(rmdir_and_freep) char *a_tmp = NULL; | |
2269 | char *b; | |
613b411c LP |
2270 | int r; |
2271 | ||
2272 | assert(id); | |
2273 | assert(tmp_dir); | |
2274 | assert(var_tmp_dir); | |
2275 | ||
56a13a49 | 2276 | r = setup_one_tmp_dir(id, "/tmp", &a, &a_tmp); |
613b411c LP |
2277 | if (r < 0) |
2278 | return r; | |
2279 | ||
56a13a49 ZJS |
2280 | r = setup_one_tmp_dir(id, "/var/tmp", &b, NULL); |
2281 | if (r < 0) | |
613b411c | 2282 | return r; |
613b411c | 2283 | |
56a13a49 ZJS |
2284 | a_tmp = mfree(a_tmp); /* avoid rmdir */ |
2285 | *tmp_dir = TAKE_PTR(a); | |
2286 | *var_tmp_dir = TAKE_PTR(b); | |
613b411c LP |
2287 | |
2288 | return 0; | |
2289 | } | |
2290 | ||
2caa38e9 | 2291 | int setup_netns(const int netns_storage_socket[static 2]) { |
613b411c | 2292 | _cleanup_close_ int netns = -1; |
3ee897d6 | 2293 | int r, q; |
613b411c LP |
2294 | |
2295 | assert(netns_storage_socket); | |
2296 | assert(netns_storage_socket[0] >= 0); | |
2297 | assert(netns_storage_socket[1] >= 0); | |
2298 | ||
2299 | /* We use the passed socketpair as a storage buffer for our | |
76cd584b LP |
2300 | * namespace reference fd. Whatever process runs this first |
2301 | * shall create a new namespace, all others should just join | |
2302 | * it. To serialize that we use a file lock on the socket | |
2303 | * pair. | |
613b411c LP |
2304 | * |
2305 | * It's a bit crazy, but hey, works great! */ | |
2306 | ||
2307 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
2308 | return -errno; | |
2309 | ||
3ee897d6 LP |
2310 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); |
2311 | if (netns == -EAGAIN) { | |
44ffcbae | 2312 | /* Nothing stored yet, so let's create a new namespace. */ |
613b411c LP |
2313 | |
2314 | if (unshare(CLONE_NEWNET) < 0) { | |
2315 | r = -errno; | |
2316 | goto fail; | |
2317 | } | |
2318 | ||
44ffcbae | 2319 | (void) loopback_setup(); |
613b411c LP |
2320 | |
2321 | netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY); | |
2322 | if (netns < 0) { | |
2323 | r = -errno; | |
2324 | goto fail; | |
2325 | } | |
2326 | ||
2327 | r = 1; | |
613b411c | 2328 | |
3ee897d6 LP |
2329 | } else if (netns < 0) { |
2330 | r = netns; | |
2331 | goto fail; | |
613b411c | 2332 | |
3ee897d6 LP |
2333 | } else { |
2334 | /* Yay, found something, so let's join the namespace */ | |
613b411c LP |
2335 | if (setns(netns, CLONE_NEWNET) < 0) { |
2336 | r = -errno; | |
2337 | goto fail; | |
2338 | } | |
2339 | ||
2340 | r = 0; | |
2341 | } | |
2342 | ||
3ee897d6 LP |
2343 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); |
2344 | if (q < 0) { | |
2345 | r = q; | |
613b411c LP |
2346 | goto fail; |
2347 | } | |
2348 | ||
2349 | fail: | |
fe048ce5 | 2350 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); |
15ae422b LP |
2351 | return r; |
2352 | } | |
417116f2 | 2353 | |
2caa38e9 | 2354 | int open_netns_path(const int netns_storage_socket[static 2], const char *path) { |
51af7fb2 LP |
2355 | _cleanup_close_ int netns = -1; |
2356 | int q, r; | |
2357 | ||
2358 | assert(netns_storage_socket); | |
2359 | assert(netns_storage_socket[0] >= 0); | |
2360 | assert(netns_storage_socket[1] >= 0); | |
2361 | assert(path); | |
2362 | ||
2363 | /* If the storage socket doesn't contain a netns fd yet, open one via the file system and store it in | |
2364 | * it. This is supposed to be called ahead of time, i.e. before setup_netns() which will allocate a | |
2365 | * new anonymous netns if needed. */ | |
2366 | ||
2367 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
2368 | return -errno; | |
2369 | ||
2370 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); | |
2371 | if (netns == -EAGAIN) { | |
2372 | /* Nothing stored yet. Open the file from the file system. */ | |
2373 | ||
2374 | netns = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC); | |
2375 | if (netns < 0) { | |
2376 | r = -errno; | |
2377 | goto fail; | |
2378 | } | |
2379 | ||
2380 | r = fd_is_network_ns(netns); | |
2381 | if (r == 0) { /* Not a netns? Refuse early. */ | |
2382 | r = -EINVAL; | |
2383 | goto fail; | |
2384 | } | |
2385 | if (r < 0 && r != -EUCLEAN) /* EUCLEAN: we don't know */ | |
2386 | goto fail; | |
2387 | ||
2388 | r = 1; | |
2389 | ||
2390 | } else if (netns < 0) { | |
2391 | r = netns; | |
2392 | goto fail; | |
2393 | } else | |
2394 | r = 0; /* Already allocated */ | |
2395 | ||
2396 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); | |
2397 | if (q < 0) { | |
2398 | r = q; | |
2399 | goto fail; | |
2400 | } | |
2401 | ||
2402 | fail: | |
2403 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); | |
2404 | return r; | |
2405 | } | |
2406 | ||
6e2d7c4f MS |
2407 | bool ns_type_supported(NamespaceType type) { |
2408 | const char *t, *ns_proc; | |
2409 | ||
0fa5b831 LP |
2410 | t = namespace_type_to_string(type); |
2411 | if (!t) /* Don't know how to translate this? Then it's not supported */ | |
6e2d7c4f MS |
2412 | return false; |
2413 | ||
6e2d7c4f | 2414 | ns_proc = strjoina("/proc/self/ns/", t); |
6e2d7c4f MS |
2415 | return access(ns_proc, F_OK) == 0; |
2416 | } | |
2417 | ||
1b8689f9 | 2418 | static const char *const protect_home_table[_PROTECT_HOME_MAX] = { |
cbc056c8 ZJS |
2419 | [PROTECT_HOME_NO] = "no", |
2420 | [PROTECT_HOME_YES] = "yes", | |
1b8689f9 | 2421 | [PROTECT_HOME_READ_ONLY] = "read-only", |
cbc056c8 | 2422 | [PROTECT_HOME_TMPFS] = "tmpfs", |
417116f2 LP |
2423 | }; |
2424 | ||
1e8c7bd5 | 2425 | DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_home, ProtectHome, PROTECT_HOME_YES); |
5e1c6154 | 2426 | |
1b8689f9 | 2427 | static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = { |
cbc056c8 ZJS |
2428 | [PROTECT_SYSTEM_NO] = "no", |
2429 | [PROTECT_SYSTEM_YES] = "yes", | |
2430 | [PROTECT_SYSTEM_FULL] = "full", | |
3f815163 | 2431 | [PROTECT_SYSTEM_STRICT] = "strict", |
1b8689f9 LP |
2432 | }; |
2433 | ||
1e8c7bd5 | 2434 | DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_system, ProtectSystem, PROTECT_SYSTEM_YES); |
03c791aa | 2435 | |
6e2d7c4f | 2436 | static const char* const namespace_type_table[] = { |
cbc056c8 | 2437 | [NAMESPACE_MOUNT] = "mnt", |
6e2d7c4f | 2438 | [NAMESPACE_CGROUP] = "cgroup", |
cbc056c8 ZJS |
2439 | [NAMESPACE_UTS] = "uts", |
2440 | [NAMESPACE_IPC] = "ipc", | |
2441 | [NAMESPACE_USER] = "user", | |
2442 | [NAMESPACE_PID] = "pid", | |
2443 | [NAMESPACE_NET] = "net", | |
6e2d7c4f MS |
2444 | }; |
2445 | ||
2446 | DEFINE_STRING_TABLE_LOOKUP(namespace_type, NamespaceType); | |
4e399953 LP |
2447 | |
2448 | static const char* const protect_proc_table[_PROTECT_PROC_MAX] = { | |
2449 | [PROTECT_PROC_DEFAULT] = "default", | |
2450 | [PROTECT_PROC_NOACCESS] = "noaccess", | |
2451 | [PROTECT_PROC_INVISIBLE] = "invisible", | |
2452 | [PROTECT_PROC_PTRACEABLE] = "ptraceable", | |
2453 | }; | |
2454 | ||
2455 | DEFINE_STRING_TABLE_LOOKUP(protect_proc, ProtectProc); | |
2456 | ||
2457 | static const char* const proc_subset_table[_PROC_SUBSET_MAX] = { | |
2458 | [PROC_SUBSET_ALL] = "all", | |
2459 | [PROC_SUBSET_PID] = "pid", | |
2460 | }; | |
2461 | ||
2462 | DEFINE_STRING_TABLE_LOOKUP(proc_subset, ProcSubset); |